Wikipedia:WikiProject AI Cleanup/Noticeboard

WikiProject AI Cleanup/Noticeboard

Shortcuts

This page is for reporting issues of AI misuse.

For issues concerning a single article, please refer to the the guide on how to deal with AI content before posting here.
You can inform an editor of a discussion involving them with {{AINB-notice}}, but this is not a requirement.
More general discussion about AI use and policies should be brought up at Wikipedia talk:WikiProject AI Cleanup.
{{AIC status}} can be used to create a hatnote keeping track of a case's status.

Start a new discussion

Archives

Archive 1

Archive 2

Archive 3

This page has archives. Topics inactive for 90 days are automatically archived 1 or more at a time by if there are more than 4.

today in large swaths of maybe-AI edits: world politics

This is another big one.

For the past year-ish, Ritwik Deuba has been adding hundreds of instances of text with consistent AI indicators (though nothing unambiguous that I've found), primarily to world politics articles. One other person has pointed this out, but they didn't respond; I left another edit on their talk page.

I'm less sure about this one than some of the others. It looks like they are probably editing the output somewhat -- although not completely enough to be detectable. It also seems like they may be using a newer LLM model and/or different prompts than most do; the older edits seem more problematic than the newer ones. In particular, most of the "reflecting the significance" AI opinion-ese is attributed as other people's opinion rather than just dangling as unattributed editorializing like it usually does, but if it's AI it still may not be an accurate summation of what those people said.

The big challenge here is that almost literally all of these edits involve very contentious geopolitical topics, and some of them have been dragged into edit wars (that don't seem to be about AI). Because of this I haven't touched them with a 1000-foot pole, besides adding the AI generated tag if the indicators were clear enough. The articles are also high-traffic enough that "just reverting" would be difficult to impossible, and might re-ignite the edit warring. So frankly I don't know what to do. Gnomingstuff (talk) 17:41, 3 October 2025 (UTC)[reply]

hmmm. do you have any diffs we can look at? I checked the edits that the other person pointed out and didn't find any hallucinations - all claims are backed by the cited source. LLMs hallucinate at high rates so I think that edit would have been significantly human reviewed. This is just one diff of course so could be missing something NicheSports (talk) 18:17, 3 October 2025 (UTC)[reply]

ok finding problems. I noticed that these three edits [1], [2], [3] were made within 15 minutes (although with significant overlap in content), without any subsequent copy-editing or typo-fixing. This one also has clear copyright violations - to the point it should be reported, which I have never done NicheSports (talk) 18:29, 3 October 2025 (UTC)[reply]

Will take another look tomorrow. I mentioned this on a few of the talk pages but some of the edits are the same paragraph added to multiple articles (which isn't necessarily a problem by itself).

For stuff like this I'm less concerned about blatant hallucinations so much as any NPOV issues/interpretations that might have been introduced by the AI. Gnomingstuff (talk) 05:58, 4 October 2025 (UTC)[reply]

It appears to be someone adding news stories with the assistance of AI. LDW5432 (talk) 15:12, 6 November 2025 (UTC)[reply]

A number of insertions from Follynomics

Came across the editor on NPP reviewing Magic fallacy, which I'm probably going to redirect somewhere, but most if not all of their major insertions appear to be unedited or minimally edited generative output. I've undid Catallactics and some others have also already been removed, for example at Antisemitic trope, thanks to Cdjp1, however there are still other edits that need to be assessed and cleaned up, such as their edits to Ludwig von Mises, which they have apparently generated 43% of current content and Gabriel Bonnot de Mably, 16%. I think those two are the major ones, but I'm mostly listing it here because I'd probably forget otherwise. Alpha3031 (t • c) 12:36, 28 October 2025 (UTC)[reply]

I noticed that this edit to Ludwig von Mises added a URL that gives a 404 error and was never archived in the Wayback Machine. Seems iffy. Stepwise Continuous Dysfunction (talk) 22:18, 30 October 2025 (UTC)[reply]

Probably going to just remove it all once I have the time to get to it. Have removed the additions to Gabriel Bonnot de Mably, added 50 words myself. I had also removed the addition to Cognitive appraisal, just forgot to note that, so pages that still need to be reviewed are Polylogism (diff, 53.1%), Destructionism, Magic fallacy (I had tagged it but didn't finish looking at it) and I think that should be all. Alpha3031 (t • c) 03:43, 5 November 2025 (UTC)[reply]

Have still been thinking of a redirect target, and I think I'd want to do it to Magical thinking if anywhere given the attestation given in the article appears to be entirely hallucinated. (Indeed, the LLM appears to have resurrected Hayek to publish the book it's attributed to in 2021, after his death in 1992 and the actual year of publication which seems to be 1988, and there are no mentions of the two words together in, AFAICT, either that book or the other two cited, The Road to Serfdom (1944) and The Counter-Revolution of Science (1952), though a few* of one or the other)

Concerningly for cleanup efforts, the IPs which have also edited this made up nonsense, which I'm provisionally calling Special:Contributions/142.245.0.0/16, though I'm not sure if a wider or narrower range could capture things better, has also inserted things into several articles, including mentions of this "magic fallacy" which fails attestation. No doubt those edits also need to be reviewed, ugh. They replied to the IP on Talk:Wall Street crash of 1929 § Unorthodox economic interpretation being presented as NPOV in opening remarks so I'm not sure if they're two people just happening to chance on the same obscure article (on an apparently made up phrase) or a case of MEAT or WP:LOUTSOCK.

* For a list of mentions that I have been able to find,

Extended content

in Road, the word "magic" does not appear to be present in the book, while "fallacy" appears only once on p. 228 as follows (emphasis mine):

After the discussions in earlier chapters it is hardly necessary to stress that these difficulties cannot be met by conferring on the various international authorities “merely” specific economic powers. The belief that this is a practical solution rests on the fallacy that economic planning is merely a technical task, which can be solved in a strictly objective manner by experts, and that the really vital things would still be left in the hands of the political authorities. [...]

Counter-Revolution has on p. 33

[...] Any knowledge which we may happen to possess about the true nature of the material thing, but which the people whose action we want to explain do not possess, is as little relevant to the explanation of their actions as our private disbelief in the efficacy of a magic charm will help us to understand the behavior of the savage who believes in it. If in investigating our contemporary society the "laws of nature" which we have to use as a datum [...]

and p. 57

[...], by their very anxiety to avoid all merely subjective elements and to confine themselves to "objective facts," to commit the mistake they are most anxious to avoid, namely that of treating as facts what are no more than vague popular theories. They thus become, when they least suspect it, the victims of the fallacy of "conceptual realism" (made familiar by A. N. Whitehead as the "fallacy of misplaced concreteness").¶

and Fatal Conceit does have "magic" in 4 different places, pp. 30,

With respect to some objects, the notion of individual property must have appeared very early, [...]. The attachment of a [good invention] to its maker [might mean that it's hard to think of selling it, and it] must accompany him even into the grave [...]. Here the fusion of inventor with ‘rightful owner’ appears, [...] story of Arthur and his sword Excalibur — a story in which the transfer of the sword came about not by human law but by a ‘higher’ law of magic or ‘the powers’.¶

74,

Indeed, to insist that all future change be just would be to demand that evolution come to a halt. [...] what would have been the effect if, at some earlier date, some magic force had been granted the power to enforce, say, some egalitarian or meritocratic creed. One soon recognises that such an event would have made the evolution of civilisation impossible. A Rawlsian world (Rawls, 1971) could thus never have become civilised: by repressing differentiation due to luck, it would have scotched most discoveries of new possibilities. [...]

98–99 (incidentally, an example of why I think the allergy of Austrians to mathematics is almost religious in nature)

But because of the delusion that macro-economics is both viable and useful (a delusion encouraged by its extensive use of mathematics, which must always impress politicians lacking any mathematical education, and which is really the nearest thing to the practice of magic that occurs among professional economists) [prices and value are dumb] explanations that vainly endeavour to account for them as ‘objective’ [...] cannot interpret the function or appreciate the indispensability of trading and markets for coordinating the productive efforts of large numbers of people.¶

and 138

How would religion have sustained beneficial customs? Customs whose beneficial effects were unperceivable by those practising them were likely to be preserved long enough to increase their selective advantage only when supported by some other strong beliefs; and some powerful supernatural or magic faiths were readily available to perform this role. [...]

but fallacy only appears on p. 27

I have no intention to commit what is often called the genetic or naturalistic fallacy. I do not claim that the results of group selection of traditions are necessarily ‘good’ — any more than [...] cockroaches, have moral value.¶
[...] whether we like it or not, without the particular traditions I have mentioned, the extended order of civilisation could not continue to exist (whereas, were cockroaches to disappear, the resulting ecological ‘disaster’ would perhaps not wreak permanent havoc on mankind); and that if we discard these traditions, out of ill-considered notions (which may indeed genuinely commit the naturalistic fallacy) [...]

(and the index, as "naturalistic fallacy", but that doesn't count since it just refers to this page here)

I think I might just go for an AFD nom actually. Someone else can decide whether they want the redirect or not. Alpha3031 (t • c) 09:06, 5 November 2025 (UTC)[reply]

The line in Polylogism about how ..the writings of Thomas Kuhn and others made relativism a mainstream doctrine is a brutal oversimplification of the philosophy of science. It doesn't reflect the given source, for sure. Furthermore, I suspect the entire "Comparison to Kuhn's incommensurability" section to be OR, whether LLM-enhanced or not.

Moreover, if we look at what the Ludwig von Mises article itself says about "polylogism", we find the following: He rejected the notion that there could be distinct sciences or truths based on race, class, or nationality, such as "Jewish science" or "German science". The given source is a book that does not mention Ludwig von Mises at all. Perhaps unsurprisingly, the text was written first and the citation added later. Stepwise Continuous Dysfunction (talk) 17:53, 6 November 2025 (UTC)[reply]

The book talks about jewish vs german science, and is a citation for such. Follynomics (talk) 18:14, 11 November 2025 (UTC)[reply]

But it doesn't talk about Ludwig von Mises, so it doesn't support the claim to which it was attached, i.e., that Ludwig von Mises rejected the notion that there could be distinct sciences or truths. Stepwise Continuous Dysfunction (talk) 01:14, 15 November 2025 (UTC)[reply]

"Magic fallacy" has been deleted. Stepwise Continuous Dysfunction (talk) 01:16, 15 November 2025 (UTC)[reply]

more serial AI rewrites

AlphaBetaGamma123 has made hundreds of quick-succession rewrites, almost all of which display the usual signs. Working through tagging them currently.

(At some point the editor was blocked for sockpuppeting reasons.) Gnomingstuff (talk) 06:04, 4 November 2025 (UTC)[reply]

Cleanup note: all edits by this account are eligible to be WP:BANREVERTed, as they were made while violating a block via socking. fifteen thousand two hundred twenty four (talk) 11:58, 16 November 2025 (UTC)[reply]

Potential LLM generations by Bechamel

Bechamel (talk · contribs) recently submitted the article 'English afternoon tea' for FA review. I came across this article in a happenstance manner, but I noticed numerous problems immediately. I suspect that an LLM was used to create this article. The smattering of dashes, the strange definitive language like Afternoon tea is served at nearly all important official, social, and sporting events – without five o’clock tea, none of these occasions would be truly British, and the presence of an odd, bulleted list concern me. A look at Bechamel's other contributions show a strange variation in writing style from article to article. I asked Bechamel about this, but he denied using an LLM. I am hoping that the veteran editors here may be able provide assistance in confirming if my assessment is correct. Yours, &c. RGloucester — ☎ 23:42, 17 November 2025 (UTC)[reply]

Discussing with editor at Talk:English afternoon tea#Problems with this article, requesting other editors here to hold off on disclosing any assessments they've made before Bechamel has a chance to respond, thank you. fifteen thousand two hundred twenty four (talk) 03:06, 18 November 2025 (UTC)[reply]

**Markdown** added at English afternoon tea [4][5] and Venetian window [6], and removed [7][8][9].

In discussion editor stated they use MS Word but not copilot. After being informed of the markdown issues they stated they couldn't rule out the possibility that Word might be using copilot mechanisms despite copilot otherwise not functioning on their Windows install.

Without evidence of greater disruptive activity, I don't think there is much else to productively do at this time. fifteen thousand two hundred twenty four (talk) 18:41, 18 November 2025 (UTC)[reply]

Thank your hard work. I have gone through the article, line by line. A combination of WP:OR of primary sources, grandiose assertions that were not supported by their citations, mistaken page numbers, strange phrasings, and close paraphrasing. This is the first time I've bothered to enter the 'AI cleanup' line of work, and I can't say it was very enjoyable. I appreciate the work everyone does here. Yours, &c. RGloucester — ☎ 08:23, 19 November 2025 (UTC)[reply]

Gyða1981

Hello! I am new to this board. I came across the article Svarfdæla saga and saw that it was tagged with containing LLM content. The user who made the suspected edits to that article, Gyða1981 [10] has also made many other edits to other articles, especially ones on lesser-known sagas, that I believe to be LLM-generated. The biggest giveaway, besides no sourcing, is that many of the edits in their edit summary contain communication intended for the LLM user (e.g. "I took the information from the Norse/Icelandic source you provided and listed the events of the saga in chronological order"). I believe this user is using an LLM to "translate" non-English text and pasting it into articles.

I have put an AI notice on all the articles in this genre that I believe Gyða1981 has pasted LLM text into. I can read Old Icelandic and can write better articles for these with correct sourcing, but I don't have very much time at the moment, and probably won't for a few months, so if anyone else can do that, I would greatly appreciate it. A few of them also probably have existing English translations of the texts in question so I can also help locate those. Also, there are a number of topics that this user has touched on that I am not as familiar with, so they should probably be looked over for LLM text as well.

I'm not sure if I should warn the user in question? I am a bit non-confrontational so maybe someone else could do that if it is appropriate. Their talk page seems to be filled with other copyright issues also. But they haven't been active on Wikipedia for a few months, and maybe notifying them will make them come back and make sneakier LLM edits, I don't know.

I'm still learning editor etiquette, so apologies if I made any mistakes. Rorb lalorb (talk) 11:01, 18 November 2025 (UTC)[reply]

Tracking subpage created and populated with articles that warrant further review or cleanup.

Accurate assessment, a warning would be appropriate but Gyða1981 has not edited since 31 August. They certainly are using an LLM to generate plot summaries, as well as to make other edits. I believe this sequence of edits at After I'm Gone is emblematic of their process:

Special:Diff/1302476101 – Added detailed plot summary based on Goodreads descriptions.
Special:Diff/1302476261 – Added detailed plot summary based on Amazon descriptions.
Special:Diff/1302476322 – Added detailed plot summary based on Amazon and Goodreads descriptions.

My favorite diff is this attack of the '''s. fifteen thousand two hundred twenty four (talk) 15:52, 18 November 2025 (UTC)[reply]

I noticed that she has also been active on iswiki so I asked her to WP:LLMDISCLOSE there, thinking it might come off more as more gentle in a smaller space: :is:Notandaspjall:Gyða1981 § LLM notkun. But I haven't heard back. --Gurkubondinn (talk) 11:15, 21 November 2025 (UTC)[reply]

I've also brought this up on :is:WP:Potturinn § Gyða1981 og LLM notkun, so this can be addressed on iswiki as well. --Gurkubondinn (talk) 14:58, 21 November 2025 (UTC)[reply]

Gyða1981 has now replied to me on iswiki, and admitted to inserting LLM generated text. In another reply to me they claimed to have used probably Grok or a comparable model, and reponded with the typical LLM-generated half-baked policy-quoting justifications for LLM use (that we keep seeing when these editors are confronted with their LLM use). --Gurkubondinn (talk) 10:15, 2 January 2026 (UTC)[reply]

Thanks for the update! Going forward, I hope we can see this editor participate in a more policy-based manner, with a focus on verifiability and neutral POV.

(teehee) Athanelar (talk) 12:39, 2 January 2026 (UTC)[reply]

I think this user should probably have WP:EXTENDED revoked, as they wouldn't have reached 500 edits without the LLM-generated edits. --Gurkubondinn (talk) 13:44, 2 January 2026 (UTC)[reply]

Unless they were using an LLM for the purpose of achieving EC (and they don't seem to have been), their editing wouldn't count as permission gaming. fifteen thousand two hundred twenty four (talk) 13:57, 2 January 2026 (UTC)[reply]

That's fair, I was mostly thinking for the purposes of protecting WP:ECR pages from LLM disruption. --Gurkubondinn (talk) 14:00, 2 January 2026 (UTC)[reply]

@Rorb lalorb You can read Old Icelandic? That's a skill I wish I had. Rare and valuable! (I'm not joking.) Cheers. David10244 (talk) 23:12, 26 November 2025 (UTC)[reply]

I'm working through the tracking page currently. This user's edits appear to be far more problematic on non-Slovene pages, so I'm focusing on those. The Icelandic saga articles are particularly egregious, which bothers me greatly as someone who appreciates Norse literature. Athanelar (talk) 18:31, 21 December 2025 (UTC)[reply]

Hello,

Thank you for the explanations and for looking into this. I am still quite new to Wikipedia and I realize now that some of my earlier edits, including my use of AI tools, did not fit well with Wikipedia’s rules for articles.

I mainly used AI to help with English wording when working from older non-English literary texts, such as Slovenian or Icelandic novellas, but I understand that this was not handled correctly in mainspace.

One reason I stepped away from editing is that I found the rules and expectations stricter and more complex than I had realized at first, and I need time to decide whether I want to stay and fully learn them. For now, I do not plan to make further edits.

I appreciate the work being done to clean up the affected articles. Gyða1981 (talk) 11:06, 3 January 2026 (UTC)[reply]

Owais Al Qarni

Owais Al Qarni (talk · contribs · global contribs · logs · block log) is an editor, also very active at bn.wikipedia.org, who has been using LLMs on English WP so prolifically since late 2023 that they gained many advanced permissions while doing so, including autopatrolled. They have created hundreds of articles according to XTools. Based on the evidence below, I think we should assume that most of this user's substantive mainspace contributions since late 2023 were LLM-assisted.

Accompanying CCI investigation

While digging into this I also discovered significant copyright violations in their userspace, flagged it for MCE89, who found more in their mainspace contributions, and a CCI request was made and a CCI investigation opened. My intuition is that the pure copyvio is primarily in their userspace or pre-LLM mainspace contributions; mainspace copyvio since late 2023 is (while apparently common) likely LLM-related.

LLM use

This editor had an intermediate level of English as of early 2023 [11][12][13] - also see [14]. By late 2023 or so they had started contributing via LLM both in and out of mainspace [15]. Some of their earlier LLM article creations were tagged by Gnomingstuff; OAQ then went to G's talk page and denied using LLMs [16]. I then independently found their edits via EF 1325 (hist · log), specifically, their recent creation of Risala Ahlus Sunnat wal Jamaat. That led to this exchange where they repeatedly prevaricated about their LLM usage. They eventually, after the CCI request was opened, admitted to having used LLMs but said they had stopped at some point [17]. I am sure their English has improved in the last few years, but I am highly skeptical of their claim about no longer using LLMs given the following, both of which this user created in the past month.

Sirat-un-Noman: they created this entire article, but see the Reviews section in particular
User:Owais_Al_Qarni/35 includes many chatbot responses about creating biographical article summaries, such as If you want, I can **continue translating the rest of his journalistic career** along with his editorial achievements and compile the full biography, covering **all his literary, translation, and journalistic contributions**.. Created at [18]

NicheSports (talk) 03:56, 30 November 2025 (UTC)[reply]

The sandbox is pretty unambiguous. I don't remember whether this was a case where I independently stumbled across multiple of their articles or only tagged the more obvious articles Gnomingstuff (talk) 06:09, 30 November 2025 (UTC)[reply]

@Gnomingstuff Based on that sandbox, and the earlier denials followed by the admission (described above), why is that user not yet blocked? They claim they are on a Wikibreak, maybe waiting for this to blow over... David10244 (talk) 06:23, 13 December 2025 (UTC)[reply]

This is a new one. I was looking at their current BN user page and noticed they have a Taliban flag there as well as a custom user box that Google translated to "This user supports the Taliban and is proud of it". Unrelated to any LLM stuff but I suppose people should be aware of this if they come back to editing NicheSports (talk) 06:04, 1 December 2025 (UTC)[reply]

Looking at the contributions, on 20 November they moved a mass of articles to "User:Owais Al Qarni/[Article Title]" with the edit summary Rewrite needed due to issues with LLM usage, which seems like a pretty flat admission in those cases I'd say. -- Cdjp1 (talk) 17:27, 1 December 2025 (UTC)[reply]

User:Bruteforce7700 - probable AI-generated text causing verification issues

I asked this user about possible LLM use last month, and they said they din't use AI. However, based on their past and current edits. I don't believe this is true. At the very least, AI summarization and/or copyediting appear to be in play. The standard WP:AISIGNS are all present, and digging into the citations attached to said signs, I have consistently found verification issues:

Source misrepresentations on Midwest Rock & Roll
Source misrepresentations on Burn (Papa Roach song)
Source misrepresentaitons on Killing Fields (song)

I stopped here, but I don't really have much reason to think the earlier stuff is better on this metric. Gnomingstuff (talk) 00:32, 2 December 2025 (UTC)[reply]

Their assertion that they don't use AI for any of [their] contributions to Wikipedia is provably untrue, utm_source=chatgpt.com is present in Special:Diff/1282797287 and Special:Diff/1281838015. This may indicate a behavioral issue. fifteen thousand two hundred twenty four (talk) 01:03, 2 December 2025 (UTC)[reply]

welp there it is

(how are you finding these?) Gnomingstuff (talk) 01:20, 2 December 2025 (UTC)[reply]

Looks like filtering the log at 1346 (hist · log) for this editor's contributions NicheSports (talk) 01:26, 2 December 2025 (UTC)[reply]

Correct, Special:AbuseLog allows searching by user. Useful for supplementing manual checks. fifteen thousand two hundred twenty four (talk) 01:50, 2 December 2025 (UTC)[reply]

looks like they're still making edits, could someone please talk to them because I just don't have it in me right now, apologies Gnomingstuff (talk) 17:06, 3 December 2025 (UTC)[reply]

Trying a different approach, I've carefully tagged all issues within an article they recently created (now draftified), and started a new topic on their talk page referencing it. Hopefully a demonstration of the issues involved will help. fifteen thousand two hundred twenty four (talk) 21:47, 3 December 2025 (UTC)[reply]

@Fifteen thousand two hundred twenty four They have continued to edit since your note on their Talk page, with (of course) no reply to you. Your diffs above are troubling. David10244 (talk) 06:26, 13 December 2025 (UTC)[reply]

Since my last, more direct message, they have only made four small and issue free edits. fifteen thousand two hundred twenty four (talk) 07:12, 13 December 2025 (UTC)[reply]

Probable LLM usage by User:OmeletteRice

@OmeletteRice: has created a large amount of pages all within minutes of each other over the past few months. Many of the pages were already moved the the draft space or deleted via deletion discussion, however ~20 articles still exist that I need help checking over. While it is completely possible that they are writing these in an external editor and copying them over, the articles written still have a bunch of signs of LLM generated text.

xtools articles created (note the time between article creations)

While I am not 100% sure on this, I do believe they are using an LLM and additional eyes would be very much helpful. – LuniZunie ツ_(talk) 17:18, 3 December 2025 (UTC)[reply]

These do seem plausibly LLM generated, yeah.

Did some spot-checking -- it's a little difficult since most/all of these sources are in Japanese. Haven't found anything that blatantly isn't mentioned by the source yet; some statements might be borderline WP:CLOP but it's hard to tell with the translation gap. Gnomingstuff (talk) 17:32, 3 December 2025 (UTC)[reply]

They have been warned several times to follow our MoS, e.g., to use Sentence Case for the name of then band they keep writing about, but they persist in using ALL CAPS FOR THE BAND NAME, because that's what the band/their marketing company/Japanese press do, which would certainly suggest at least a lot of copy-pasting. Bastun^{Ėġáḍβáś₮ŭŃ!} 17:49, 3 December 2025 (UTC)[reply]

Considering a lot of these are in Japanese, I'll look into this case. AlphaBetaGamma (Talk/report any mistakes here) 10:05, 9 December 2025 (UTC)[reply]

Couldn't figure out a good summary to this. Surviving articles doesn't look too bad after a quick glance. Notability concerns and tone/promotion issues left and right, however/ AlphaBetaGamma (Talk/report any mistakes here) 10:18, 9 December 2025 (UTC)[reply]

@AlphaBetaGamma Agreed, I would also note that this has now been discussed with the user here. – LuniZunie ツ_(talk) 13:12, 9 December 2025 (UTC)[reply]

User:Adamsecretxx

Adamsecretxx (talk · contribs)

Most of the edits that have been made show signs of LLM (most significantly the bolding, and some sources not linking to what they say they link to), but not enough for G15; additionally, most of the articles that have been created seem to be non-notable, but in that frustrating boundary region where it's not obvious whether it seems non-notable because of the LLM or because it is. (Or maybe I'm being too lenient here.) User has also made significant changes to Victor Mayer, a preexisting article, and made half of it look AI generated. Would appreciate more eyes on the contribs. Fermiboson (talk) 22:00, 3 December 2025 (UTC)[reply]

Agreed that most/all of their substantive mainspace contributions seem to be unreviewed LLM. Albert Mayer (Davos) is G15-able (looks like you tagged it for A7). I think Marie Fabergé can be G15'd as well, and it is teeming with WP:AISIGNS. Also see Special:Diff/1322820628: I wonder about this editor's English level. I will drop them a note on their talk page. We should create a tracking page as well NicheSports (talk) 23:10, 3 December 2025 (UTC)[reply]

Thank you for the comments. I´ll be more mindful firm now on to explain my edits and new page proposals. Adam Adamsecretxx (talk) 00:44, 6 December 2025 (UTC)[reply]

+1, as the kids say. I draftified one of their creations a couple of weeks ago for a possible combination of LLM and CIR issues, and now their talk page has lit up with many similar notices and no response from this user. ClaudineChionh (she/her · talk · email · global) 13:48, 4 December 2025 (UTC)[reply]

Hi, thank you for the messages.

I´m admitting that I had LLM check up on my texts. I did not know thats not allowed. the research and the original texts were mine. I¨m also not sure what is noticeabale and what not.

MARIE FABERGE For example Marie Fabergé is the wife of one of the most prominent jewellers ...so is she noticeable or not? The entire lineage of Fabergé is written about in basically every monograph about carl Fabergé. It is not up to me to decide.

I do have knack to research topics that are NOT extremely published, so that might be the problem, plus that I used ai to check up on my form and content, I´ll update how I work.

REUCHLINHAUS I did edit one page now. Reuchlinhaus, I hope it looks clean. The German wikipedia already has a page on that subject, so at least in germany it seems to be noticeable, and I kept it very close to what was in the German wikipedia. Thanks for checking my writings and for being lenient with me as an on and off user. Adam Adamsecretxx (talk) 23:27, 4 December 2025 (UTC)[reply]

I¨m also not sure what is noticeabale and what not – The relevant policy is WP:NOTABILITY, and the section to pay the most attention to is the one on the general notability guidelines. Do note that the English and German Wikipedias have different standards for notability, so a topic that is notable on one project may not be on the other. (You may wish to visit the Introduction, which can help get you up-to-speed on what is expected from editors on enwiki.)

As a more general note about large language models, they are ill-suited to assist with editing. LLMs are prone to hallucinations, have issues maintaining a neutral tone, and often produce prose incompatible with the manual of style. As these models are predictive, they can predict incorrectly even for simple prompts or requests. Thank you for discussing this with us. fifteen thousand two hundred twenty four (talk) 00:31, 5 December 2025 (UTC)[reply]

Yes. I study the guidelines now for the different countries and it becomes more clear. It helps to be sure about what generally all the rules are, albeit jn detail it iS about a group effort. Adamsecretxx (talk) 00:43, 6 December 2025 (UTC)[reply]

Your article on Maria Fabergé contains sources that are entirely hallucinated. The ancestry link goes nowhere and the book "Gustav Fabergé and his Dresden Years" doesn't even seem to exist exist; the only results for it on google are the Wikipedia articles for the Fabergés, and gbooks turns up nothing. Your fourth {{cite web}} reference doesn't even have a URL specified. Athanelar (talk) 01:38, 10 December 2025 (UTC)[reply]

Hello, Yes the text is largely based on that book with the recent research. It¨s not hallucinated. the link does not go to a real page and I will see that I can find a link that works. Adamsecretxx (talk) 02:06, 10 December 2025 (UTC)[reply]

The text is based on what book? The book "Gustav Fabergé and his Dresden Years" doesn't seem to exist. I can't find any evidence of it whatsoever. Can you show some indication that it exists?

How did you end up adding a link to an Ancestry page that doesn't exist? Athanelar (talk) 02:10, 10 December 2025 (UTC)[reply]

For future reference, the citation in question is:

<ref name="Skurlov">{{Cite book |last=Skurlov |first=Valentin |title=Gustav Fabergé and his Dresden Years |publisher=Igor Carl Fabergé Foundation |location=Moscow |year=2018}}</ref>

Could you please provide an ISBN or other similar identifier? The title does not appear to exist. fifteen thousand two hundred twenty four (talk) 02:15, 10 December 2025 (UTC)[reply]

I suggested above that that article could probably be G15'd, but after seeing this source analysis that probably is now definitely. You want to take it? NicheSports (talk) 02:09, 10 December 2025 (UTC)[reply]

Take a look, I've already tagged it. Athanelar (talk) 02:10, 10 December 2025 (UTC)[reply]

Nice. Thanks! NicheSports (talk) 02:12, 10 December 2025 (UTC)[reply]

It looks like they're still at it (see these edits to Jardines del Bosque and Jardines del Pedregal) despite stating they would stop using LLMs on their talk page. Should this be brought to ANI? Zygmeyer (talk) 21:39, 16 December 2025 (UTC)[reply]

I think so, yes NicheSports (talk) 21:42, 16 December 2025 (UTC)[reply]

I've never done this before, is there a guide for the procedures I should follow or should I just go for it? Zygmeyer (talk) 22:16, 16 December 2025 (UTC)[reply]

There are instructions at the top of the noticeboard. The most important point is to present relevant diffs, which have already been compiled in this thread, the tracking page, and the user talk page. You also need to ensure that you notify the user you're reporting. Twinkle can help you follow all the necessary steps. ClaudineChionh (she/her · talk · email · global) 22:22, 16 December 2025 (UTC)[reply]

I misspoke slightly – Twinkle won't fill out a whole report, but it can simplify the process of notifying users. ClaudineChionh (she/her · talk · email · global) 22:44, 16 December 2025 (UTC)[reply]

Hi, I see your conversations about my account. What exactly did I do at the pages Jardines del Pedregal and Jardines del Bosque that goes against Wikipedia rules? - Adamsecretxx (talk) 22:49, 16 December 2025 (UTC)[reply]

@Adamsecretxx, did you use a large language model/AI chatbot to generate new text for those two articles? I can't speak for @Zygmeyer, but those two edits have a different style and level of English fluency compared to your responses on talk pages, which makes me suspect AI use. ClaudineChionh (she/her · talk · email · global) 23:00, 16 December 2025 (UTC)[reply]

Thanks for asking directly. Yes, I did use an AI/LLM tool at times to help rephrase or polish text, especially because English is not my first language. I now understand that this is discouraged for mainspace editing because it can create problems with tone, accuracy, and sourcing.

I’m stopping that practice for Wikipedia edits. I’m reviewing my recent contributions, reverting or trimming anything questionable, and moving any further work to draft or sandbox so it can be checked carefully before publication. Adamsecretxx (talk) 23:13, 16 December 2025 (UTC)[reply]

I was going to respond to this, but now that a report has been made on Administrators' noticeboard/Incidents, let's continue the discussion there. ClaudineChionh (she/her · talk · email · global) 02:20, 17 December 2025 (UTC)[reply]

DJ Sturm at Estonia

The LLM edits by DJ Sturm (talk · contribs) to the Estonia article need an edit-by-edit tracker for this special case, since the LLM edit reverts have to be integrated with subsequent revisions. I don't think this deserves notifying DJ Sturm. –LaundryPizza03 (d c̄) 02:33, 4 December 2025 (UTC)[reply]

To be fair, notifying them would be somewhat pointless since they've been infed'd sitewide for their conduct on commons. Though I'd question the "deserves notifying" phrasing in general. --Gurkubondinn (talk) 11:48, 4 December 2025 (UTC)[reply]

Edits at Pop music

I'm concerned about the edits of User:User7312549 at Pop music. I have twice deleted the edits, but a second opinion would be great. Thank you! Magnolia677 (talk) 22:41, 4 December 2025 (UTC)[reply]

Not sure on AI use here, but the source-to-text integrity is poor, and the choice of sources is also not great:

In the late 1990s and early 2000s the digital distribution of music became increasingly popular, platforms like CD Baby and Tunecore made it easier for independent artists to release and distribute music without a major label - The source it is cited to does not mention CD Baby/CDBaby or TuneCore. The source also doesn't mention any of the 1990s, "nineties," any 199X year, etc.

During the early to mid 2000s streaming platforms became popularized with apps like Pandora and Spotify, pushing artists to prioritize singles over full albums. The source (paywalled but accessible via Wikipedia Library) does not mention Pandora (which was pretty minor in terms of impact anyway). The sentence here also does a poor job of explaining why streaming platforms encourage artists to prioritize singles over full albums; skimming the source there does seem to be potential information about that.

Pop music shifted towards shorter songs, with catchy hooks, and shorter attention grabbing intros, optimized for replay. -- Source-to-text integrity OK although there is probably a better source than The Washington Times. "Catchy hooks" is also a bit sus - when didn't pop songs have catchy hooks?

Advances in digital audio workstations (DAWs) and home recording equipment made music production more accessible, allowing independent and emerging artists to create commercially ready tracks without relying on traditional studio settings. Probably verifiable but not backed up by the source it's cited to, mostly because the source appears to be someone's crappy business-school assignment (Choose a company that provides an interesting example of how digital transformation has created opportunities or challenges for business and operating models.)

Social media became increasingly popular during this time with platforms like MySpace and YouTube allowing artists to share their music and connect to fans more directly. This created a shift from traditional promotion methods to a more direct and interactive model. - Cited to an AI slop advertorial blog post that doesn't mention MySpace or YouTube. The second sentence is also redundant -- it didn't "create" a shift, it is the shift.

Research on contemporary songwriting practices indicates that platform standards shape the structure and format of pop songs, with elements adjusted to fit algorithmic playlist norms, including shorter intros, earlier hooks, and clearly defined song sections. This is probably verifiable but not corroborated by the source it is cited to. "Research" is also WP:WEASELWORD-ing.

In the late 2010s-early 2020s, TikTok became a powerful tool for fan engagment and viral marketing. Challenges and memes made music trends and hit go viral giving rise to rapid chart success. In 2024, 84% of songs that entered the Billboard Global 200 chart initially went viral on TikTok, indicating the platform's role in influencing global pop music chart performance. - The statistic is corroborated by the source, but not the stuff before it; furthermore, this whole sentence is promotional in tone. The last clause is also redundant captain obvious stuff -- and does sound very much like AI.

In the words of Anthony Fantano: not good. Gnomingstuff (talk) 02:02, 5 December 2025 (UTC)[reply]

Per Gnomingstuff's analysis, yeah that's AI. Because this is a WikiEd situation you should contact the instructor and/or User:Brianda (Wiki Ed), who actually reverts a lot of this content herself. I'm seeing a ton of this now bc of the end of the semester. NicheSports (talk) 02:19, 5 December 2025 (UTC)[reply]

Thank you all very much for your help. Magnolia677 (talk) 11:20, 5 December 2025 (UTC)[reply]

Please feel free to ping myself or @Ian (Wiki Ed), when you encounter work that is not suitable for Wikipedia from our student editors, especially when it might be AI. Brianda (Wiki Ed) (talk) 22:57, 8 December 2025 (UTC)[reply]

Student who used LLM to write drafts on articles that already exist

Luis jimenez is an LLM-generated article that is up for CSD A10, as its subject is already covered by Luis Jiménez (sculptor). Creator Ryan162j (talk · contribs), who is enrolled in a student program, also created numerous subpages in their userspace, including two more drafts on Luis Jiménez and one on the Blue Mustang. See Wikipedia:Miscellany for deletion/User:Ryan162j/Luis jimenez. I am not sure how to proceed on other subpages.

Other users at Wikipedia:Wiki Ed/California State University, San Bernardino/Latinx Art and the Politics of Presence (Fall 2025), e.g. User:Lamiy23/Mahmoud Mokhtar, also used LLMs for their work on subjects that are already covered here. –LaundryPizza03 (d c̄) 07:38, 5 December 2025 (UTC)[reply]

not sure this is fully AI, at the very least it has been [poorly] human-edited. ("Luis Jimenez is just a sculptor" is pretty funny, even refreshing, in context.)

that said, probably the same advice as above, contact the instructor and/or wiki ed representatives. Gnomingstuff (talk) 14:28, 5 December 2025 (UTC)[reply]

Quickdrew and possible AI hoax edits to contentious US politics topics

First off, all of this user's edits display the gamut of WP:AISIGNS, this is the baseline we're working off and should be assumed. What elevates this to an actual noticeboard post, though, is two things.

One, most of their edits involve the Trump administration, QAnon, etc., all very contentious topics we don't want to get things wrong in. (also CAN bus for some reason) There are only a few articles they've added, but they've added a lot of text.

Two, their version of the Mar-a-Lago Accord article has severe hallucination issues bordering on WP:HOAX. I nominated it for AfD but tl;dr: the draft calls it the "Margo Largo Accord" and discusses it as if it had actually been proposed and gotten reactions and analysis, which it hasn't. Even assuming good faith, this is very bad. Obviously they're not doing much review of their own output, and given that the Mar-a-Lago Accord article was largely unchanged until now it seems like others are at least sometimes taking that output at face value. Gnomingstuff (talk) 20:41, 6 December 2025 (UTC)[reply]

This actually led me down a rabbit hole of the Foreign policy of the second Donald Trump administration article as a whole, which this user was a regular on.

I just did a massive trim of the Spain section there, because it had a lot of AI generated content, source-to-text discrepancies and verifiability issues. This combined with that user's activity there makes me think this article needs a fine-toothed comb; I intend to do it a bit myself, but it's about time for me to get off Wikipedia for the moment, so I thought I'd drop this here in case anybody else wants to give it a look over in the meantime. Athanelar (talk) 05:18, 8 December 2025 (UTC)[reply]

yeah I've looked at that article before, I suspect a lot of it is either AI generated or AI edited but there are just so many edits to slog through Gnomingstuff (talk) 14:38, 8 December 2025 (UTC)[reply]

142 edits to mainspace across 9 articles mostly related to post-1992 American politics and conspiracy. Cleanup requested. Foreign policy of the Biden administration seems to have been hit by many other LLM users, and needs immediate life support. AlphaBetaGamma (Talk/report any mistakes here) 10:01, 9 December 2025 (UTC)[reply]

Ekabhishek

I was surprised to find all nonexistent links in this large addition. The editor Ekabhishek (talk · contribs) reverted themselves after some nagging and have stopped editing for the last couple of weeks. I found more instances of llm generated text added wholesale by them.

Going back in their contributions list up to July 2025 and looking at large edits, I've boldly (they are an admin) removed their edits which weren't backed up by the refs, from Meenakshi Jain, Mrs (film) and Dastak (1970 film) as original research. But I keep finding more - like [this edit] where the last ref link is a page not found - as I go further back. Legospy (talk) 06:41, 7 December 2025 (UTC)[reply]

Yikes, that first example is textbook AI text. I've seen NPPs and autopatrolled users throw LLM text around, but admins? That's new. I'm going to see how deep is this. AlphaBetaGamma (Talk/report any mistakes here) 03:43, 8 December 2025 (UTC)[reply]

Yeah that sure has a few WP:AISIGNS. And it had a bunch of broken links? That is a G15-level of proof of unreviewed LLM content. I also see that Ekabhishek chose to not answer three questions on their talk page from Legospy about LLM usage. @AlphaBetaGamma I will add an AINB notice to their talk page and tell them we need to hear from them. Do you want me to wait until you've dug more? NicheSports (talk) 04:04, 8 December 2025 (UTC)[reply]

Looks like they haven't edited since November 18 so no rush on the AINB notice. I'll wait until you've looked into this. NicheSports (talk) 04:23, 8 December 2025 (UTC)[reply]

Would it not be wise to bring it up at WP:AN? I would imagine an admin editing disruptively like this should at least warrant a reconsidering of their access to the tools. Athanelar (talk) 04:44, 8 December 2025 (UTC)[reply]

They haven't edited in weeks so there's no rush imo. Let's look into their edits first? Also too early to talk about recall. NicheSports (talk) 05:04, 8 December 2025 (UTC)[reply]

I know this is a blatant assumption of bad faith that should usually be avoided, but if they simply return to editing and fly under the radar with their edits again, wouldn't the damage be severe? I can't feel safe until they give us a non-dodgy explanation/declaration. I really do not want an another recall drama (it has already depressed me a lot) unless the admin in question causes a WP:ANI level of disruption and refuses to communicate.

Speaking of digging through, I was seeing if I needed to blast off any faulty articles, but I don't think there is an urgent need compared to their edits. they have made around 11 articles in 2025, and all of their other creations predate ChatGPT. Stuff like Bina Ramani had some eye-raising issues with promotional words (pionneering with no cited source actually saying that word). However, their 11 creations in 2025 probably needs some duct taping, so cleanup is requested. AlphaBetaGamma (Talk/report any mistakes here) 05:45, 8 December 2025 (UTC)[reply]

Judging by their creation activity, we might need to scan every one of their mainspace edits after the admin's return from inactivity in 2024ish, which is easily over a thousand. Holy hell... AlphaBetaGamma (Talk/report any mistakes here) 05:52, 8 December 2025 (UTC)[reply]

Sigh. Thanks for looking into this. I agree with Athanelar that this probably needs to go to AN if Ekabhishek doesn't return and address these issues soon. If we go there, we should try to quantify the extent of the LLM misuse and bring representative diffs - any chance you can keep working on that? Added: Fifteen's diffs below look compelling. We could also ping an admin to get advice. Newslinger comes to mind. And if at any point Ekabhishek returns to editing without engaging with us I think we should go to AN right away. NicheSports (talk) 06:34, 8 December 2025 (UTC)[reply]

I've notified Ekabhishek of this discussion at User talk:Ekabhishek § AI cleanup noticeboard discussion. — Newslinger talk 06:40, 11 December 2025 (UTC)[reply]

"Pioneering" is probably a rendition of descriptions like this: "What really makes Bina Ramani's transformation of Hauz Khas Village so remarkable and memorable is that she did much more than just reviving a neighbourhood, she birthed the vibe of the place. She turned ruins into rendezvous spots, history into heritage, and then made the heritage turn into a social movement. It wasn’t long before the area became synonymous with indie fashion, experimental art, and late-night conversations under fairy lights." (The Godmother of Hauz Khas And Its Transformation Of The '80s (cited in th Bina Ramani article))

Another article that uses the word "pioneer" to describe Bina Ramani is this one.

The first case Legospy cited and the examples fifteen thousand two hundred twenty four analyzed sound like llm/hallucination though. Deamonpen (talk) 08:16, 8 December 2025 (UTC)[reply]

Special:Diff/1318901237 – both quote=s appear to be hallucinations
Special:Diff/1307138307 – contains markdown-style footnote refs: [^3][^4]
Special:Diff/1295888253 – The Sabarmati Report And Hindi Journalism: A Political Narrative Disguised As Cinema does not exist, The Sabarmati Report movie review: Refreshing and fast-paced does not exist, The Sabarmati Report Review: Vikrant Massey shines in this ideological drama does not exist, The Sabarmati Report Review: A riveting but one-sided narrative does not exist (here is what the homepage of "thebrew.in" looked like in May 2024 when the supposed article was published).

This was a surface-level look. I've not included any snippets of LLM prose which often overstep the bounds of WP:NPOV, nor utm_source indicators. fifteen thousand two hundred twenty four (talk) 04:36, 8 December 2025 (UTC)[reply]

They're unambiguously using AI for some things -- here is an early sandbox edit that contains attributableIndex json all over the place Gnomingstuff (talk) 22:13, 8 December 2025 (UTC)[reply]

In some recent additions, I may have used ai for copy editing and improvement, but rewritten to avoid such a scenario, though I admit some mistakes would have crept in. So if you feel like it is too much in some places, please go ahead and remove or rewrite, no issues. Since we are all here to BUILD wikipedia and in good faith. I will also try to rewrite whatever it is possible next. Thanks --Ekabhishek^talk 14:02, 17 December 2025 (UTC)[reply]

I may have – Why the uncertainty? An LLM has definitively performed more than copy editing as evident by multiple hallucinations. Regardless, LLMs are unsuited for copyediting, they are predictive models and can predict wrong even with very simple prompts, leading to the introduction of hallucinations, biases, unencyclopedic prose, and non-neutral constructions.

Acting in good faith is half the equation, competence is the other half and requires editors understand the tools they use to ensure that, at a minimum, outright falsifications are not blindly copied-and-pasted into the project. fifteen thousand two hundred twenty four (talk) 00:09, 18 December 2025 (UTC)[reply]

Multinational Force – Ukraine

The main author, D'Lisye (talk · contribs · deleted contribs · logs · filter log · block user · block log), of Multinational Force – Ukraine has been constructively engaging in discussion in relation to LLM use. However, the earlier addition of a Microsoft Copilot URL and addition of a Google Gemini URL as sources shows that at least earlier, the user severely lacked understanding of WP:VERIFIABILITY and of what LLMs actually are. I admit that I was flabbergasted. Anyway, I reverted both of those edits.

My recommendation was a voluntary de facto WP:TNT of Multinational Force – Ukraine by the main author; my suggestion doesn't seem to have been accepted. Just now in browsing LLM discussions, I see that some common LLM tracers are present in the current version of the article. Just cleaning up obvious tracers seems pointless to me, since then all the more nuanced problems will remain, requiring much more editing energy in making corrections than would be needed to rewrite from scratch (by human-summarising key points in the sources, per all the regular Wikipedia policies and guidelines).

I suggest a clean up if someone is willing to invest the huge of amount of energy needed to do so, or do a soft WP:TNT with a stub rewrite, or propose a formal WP:AFD for WP:TNT (the topic itself is clearly notable). Boud (talk) 16:15, 9 December 2025 (UTC)[reply]

It's possible that the editor may actually accept a rebuild from a stub, i.e. a soft WP:TNT. Let's see ... Boud (talk) 17:23, 9 December 2025 (UTC)[reply]

The editor accepted to restart, but is restoring a lot of LLM-style material. Boud (talk) 20:46, 9 December 2025 (UTC)[reply]

Three years of bad AI copyedits by User:Kofi Meija

This user has made hundreds of copy edits between 2023-2025 (several of them prompted by the Newcomer Edits gamification) and is still active doing that. These edits appear to be AI-generated, based on the usual WP:AISIGNS that recur throughout. (Their talk page posts also seem like "Certainly! You're absolutely right"-type AI responses.)

Unsurprisingly, many of these changes display the usual problems with AI copyedits -- namely, claiming they are making the article neutral but actually making it promotional, claiming they are just copyedits but actually change meaning or introduce new content, or just generally making things wordy and awkward.

Here's a quick spot-check of some of those copyedits (I have only looked at a few) to demonstrate. I've left out citations and any sentences/paragraphs that are identical/only superficially changed.

Blue = introduced new information, removed information for unclear reasons, or changed meaning
Green = introduced puffery
Orange = introduced clunky, wordy, or otherwise bad phrasing

Extended content

Tower 25 -- edit made June 2023, "Grammatical fixes throughout the article and rephrasing the article to have a more neutral and clear tone."

Previous text: The ground floor, the mezzanine areas, and the six floors above them are offices of Ernst & Young (EY). The next seven floors contain apartments overlooking the whole capital. The penthouse occupies two top floors. Its design resembles traditional Cypriot architecture, and contains a swimming pool. Most of the apartments, including the penthouse, were sold before construction commenced. The building cost approximately €25 million and was finished by early 2013.

A full-length balcony on each floor offers views of the historic city and walls, also allowing natural light in the living spaces and offices below. The varying width and depth of balconies on each floor create a natural image as well as the illusion that the building is not static but actually breathing. The other facades feature a seemingly random pattern of square voids serving as windows as well as openings for natural ventilation needed due to the hot climate of the city. On the south-facing facade, which receives the most sunlight, a series of balconies was created that span the width of the building.

The building's location faces the Venetian fortifications of the city and is next to the new Eleftheria Square redesigned by architect Zaha Hadid. It is also close to the main retailing avenues of the city, Makariou Avenue and Themistokli Dervi Avenue. The building itself is meant to attract more developments on the ring avenue around the walls on which it is located.

AI edit: The ground floor, mezzanine areas, and six floors above them are occupied by offices belonging to Ernst & Young (EY). The subsequent seven floors consist of apartments that offer panoramic views of the entire capital city. The penthouse, spanning two top floors, showcases a design inspired by traditional Cypriot architecture and features a swimming pool. A significant portion of the apartments, including the penthouse, had already been sold prior to the commencement of construction. The total cost of the building was approximately €25 million, and its construction was completed by early 2013.

Each floor of the building boasts a full-length balcony, offering captivating views of the historic city and its walls. These balconies not only provide an opportunity for natural light to illuminate the living spaces and offices below but also contribute to the overall aesthetic appeal of the structure. The balconies on each floor vary in width and depth, creating a sense of organic fluidity and imparting the illusion that the building is in a state of dynamic movement. The other facades of the building exhibit a pattern of square voids that serve as both windows and openings for natural ventilation, catering to the hot climate of the city. Particularly on the sun-facing southern facade, a series of balconies extend across the entire width of the building, effectively addressing the need for shade and facilitating a pleasant outdoor experience.

The building is situated in a prominent location that overlooks the historic Venetian fortifications of the city. Adjacent to the newly redesigned Eleftheria Square, envisioned by architect Zaha Hadid, it benefits from its proximity to the city's major commercial thoroughfares, namely Makariou Avenue and Themistokli Dervi Avenue. The strategic positioning of the building is intended to stimulate further development along the ring avenue encircling the city walls.

. . .

Lisa Ross (photographer) -- edit made November 2024, "Cleaned up for a more neutral tone."

Previous text: Lisa Ross is a photographer. [largely unchanged text omitted]

AI edit: Lisa Ross is an American photographer known for her work exploring themes of cultural heritage and spirituality. [largely unchanged text omitted]

. . .

CogPrints -- edit made February 2025, "Revised article for a more neutral tone."

Previous text: CogPrints is moderated by Stevan Harnad. The archive was launched in 1997 and now contains over 2000 freely downloadable articles. Some cite CogPrints, along with the physics archive arXiv as evidence that the author self-archiving model of Open Access can work—although under the influence of the Open Archives Initiative and its OAI-PMH, the emphasis in self-archiving has since moved away from such central repositories in the direction of distributed self-archiving in Institutional Repositories. CogPrints was first made OAI-compliant, and then the software was converted into the EPrints software at the University of Southampton by Rob Tansley who then went on to design DSpace. EPrints is now maintained by Christopher Gutteridge at Southampton.

AI edit: The archive was launched in 1997 and is moderated by Stevan Harnad. As of 2025, it contains over 4,000 freely downloadable articles. CogPrints has been referenced alongside the physics archive arXiv as an example of the author self-archiving model within open access publishing. Over time, under the influence of the Open Archives Initiative and its OAI-PMH, the focus of self-archiving has shifted from centralized repositories like CogPrints to distributed self-archiving in institutional repositories. CogPrints was among the first repositories to adopt OAI compliance. Its software was later converted into EPrints at the University of Southampton by Rob Tansley, who later contributed to the development of DSpace. EPrints is currently maintained by Christopher Gutteridge at Southampton.

Anyway let me know thoughts. There are so many of these going back to like 2023. Gnomingstuff (talk) 01:09, 12 December 2025 (UTC)[reply]

I've recently seen a lot of adding of different |language= parameters to citations in what look like AI edits, see [19], not sure if it's new or something I've not noticed before. CMD (talk) 02:00, 12 December 2025 (UTC)[reply]

I was skeptical, but I ran the numbers...

Word	Change	Human	AI	Chi square signif.
\|language=en}}</ref><ref>{{Cite	5473.21	1.86	103.41	True
\|language=en-US}}</ref>	4348.35	6.49	288.88	True
\|language=English	1280.52	2.78	38.42	True
\|language=en}}</ref>	1181.10	33.40	427.87	True
\|language=en	1143.71	17.16	213.46	True
\|language=en}}</ref><ref	794.78	1.86	16.60	True

So... yep that appears to be a thing. (Unless there's some kind of wizard change introduced after 2022 to automatically add it, that the human dataset wouldn't have.) Gnomingstuff (talk) 02:25, 12 December 2025 (UTC)[reply]

Have been going through these. The problematic stuff seems to start around June 2023. As expected, quality varies widely on these, they are all a mixed bag of acceptable to pointless to actively detrimental to bad bad, but I'm not even 2 weeks' worth of edits in and I've already found one that's pretty bad: "rephrasing the article to have a more neutral and clear tone" by changing "She did not support Donald Trump" to "She withdrew her support for Donald Trump" (implying that she did, at one point, support him) out of fucking nowhere. I even went back to the original version of the article to see whether it was maybe restoring some old text; it wasn't. Living person by the way.

How would I make a tracking page for these? Gnomingstuff (talk) 04:26, 14 December 2025 (UTC)[reply]

They've edited 3174 different pages since June 2023, not great.

How would I make a tracking page for these? – Either use the AINB-helper script to create the subpage (not advised in this case), or create it manually. When manually creating large lists, I use CCI's contribution surveyor (uncheck minor, set bytes to -999999, in this instance set a date range too) to fetch all edited pages, use regex to format the lines into {{AIC article row}}s, wrap the rows in {{AIC article list}}, then paste and preview the result in the editor and perform some manual checks and refinement if able.

Once the subpage is created, populate the tracking_subpage= parameter of {{AIC status}}. fifteen thousand two hundred twenty four (talk) 06:28, 14 December 2025 (UTC)[reply]

welp. made the page; it starts in mid-March as this is when the "rephrasing" edits start. have not filtered it yet beyond that — Preceding unsigned comment added by Gnomingstuff (talk • contribs) 03:40, 15 December 2025 (UTC)[reply]

welp indeed, so many entries it hits the transclusion limit if the templates are used. I've instead manually added the count to {{AIC status}}. fifteen thousand two hundred twenty four (talk) 08:19, 15 December 2025 (UTC)[reply]

The good news is that a lot of them are probably quick fixes and/or irrelevant, have not done any manual culling yet Gnomingstuff (talk) 22:46, 15 December 2025 (UTC)[reply]

Has nobody messaged them? They were back editing yesterday Kowal2701 (talk) 15:57, 22 December 2025 (UTC)[reply]

I haven't. I don't have the patience to deal with that right now, apologies. Gnomingstuff (talk) 08:00, 28 December 2025 (UTC)[reply]

No worries, I messaged them. They've promised not to use LLMs anymore (though the comment seemed AI-generated) Kowal2701 (talk) 08:15, 28 December 2025 (UTC)[reply]

I tried to talk with them but all their comments seem AI-generated. They said they’ll stop using it and edit at Twi wiki. Gnomingstuff, would it be possible to keep an eye on their edits and lmk if any are unambiguously AI? Kowal2701 (talk) 15:36, 6 January 2026 (UTC)[reply]

I'm not Gnomingstuff, but I can keep an eye on things for a few months. fifteen thousand two hundred twenty four (talk) 15:47, 6 January 2026 (UTC)[reply]

Hi, sorry for not going through more of the edits, other stuff has been going on and I haven't gotten back to this. I don't really have time to specifically monitor their new edits, apologies. Gnomingstuff (talk) 15:49, 6 January 2026 (UTC)[reply]

No worries, thanks 15224 Kowal2701 (talk) 15:50, 6 January 2026 (UTC)[reply]

I wanted to bring up this edit where they Changed the background color to a more suitable one. (from red to yellow) a few years back. I reverted it about a month later because I didn't think it was an improvement. It being AI edit didn't even cross my mind at the time, and I had no way of knowing. What I've seen here, however, brings a lot into perspective. XtraJovial (talk • contribs) 18:01, 31 December 2025 (UTC)[reply]

User:~2025-31172-04

Could someone please have a look at User:~2025-31172-04. Enormous edits with telltale signs of AI. Multiple cut-and-paste warnings. Thank you! Magnolia677 (talk) 11:22, 16 December 2025 (UTC)[reply]

Win8x (talk · contribs) has put in a lot of work cleaning up this user's attribution issues. I left a new section on their talk page. I am not optimistic as the temporary account has zero edits to Talk: or User talk: spaces. -- Recon rabbit 17:02, 16 December 2025 (UTC)[reply]

Checks for machine translation might be in order. I went through their edits quickly to attribute them, didn't read them all. There's French, German, Portuguese, Spanish translations. Not saying it's not possible, but there's few people who speak 5 languages! win8x (talk) 17:07, 16 December 2025 (UTC)[reply]

They have continued editing, so I have opened up a thread at ANI. I don't know if this is "obvious spam" that could be presented at AIV as was done below. -- Recon rabbit 14:29, 18 December 2025 (UTC)[reply]

Created a tracking subpage for the larger edits. Zygmeyer (talk) 04:28, 17 December 2025 (UTC)[reply]

I ran into a similar TA the other day: ~2025-40162-16 (talk · contribs). Reported to AIV and have cleaned up (most of) the mess. --Gurkubondinn (talk) 23:26, 17 December 2025 (UTC)[reply]

User:MaynardClark

Fairly prolific with edits, having the usual WP:AISIGNS all over the place going back to 2023; but what elevates this to an actual noticeboard post is the talk page comment here, User_talk:MaynardClark#ChatGPT, specifically the last paragraph which suggests that review isn't really being done. Gnomingstuff (talk) 23:20, 16 December 2025 (UTC)[reply]

Possibly AI-generated page

Wikipedia:Wiki Ed/Minnesota State University Mankato/Black Feminist Thought (Fall 2024) contains a lot of AI vocabulary. I could probably list them all:

"delve"
"rich tapestry"
"intricate interplay"
"profound connections"
"broader cultural narratives"
"panorama"
"nuanced"

I'm not too knowledgeable on Wikipedia, so I don't really know how WikiEd works. ~2025-31416-56 (talk) 14:07, 17 December 2025 (UTC)[reply]

Yeah this seems like AI. I could be wrong but I think these are syllabi write by professors (which, uh......) so not sure whether changing them would be appropriate. Gnomingstuff (talk) 16:22, 17 December 2025 (UTC)[reply]

Maybe report it to someone who can properly deal with it? If this was truly written by a college professor, it might need to be dealt with outside of Wikipedia as well. ~2025-31416-56 (talk) 17:18, 17 December 2025 (UTC)[reply]

Decided to take it to ANI. Hopefully something can be worked out there. ~2025-31416-56 (talk) 17:41, 17 December 2025 (UTC)[reply]

Hmm. The course is from Fall 2024 and the instructor has a total of 2 edits, both from over a year ago. I don't think this needs to be at ANI. Wiki Ed has a lot of problems, I don't really see any action being taken here. NicheSports (talk) 18:16, 17 December 2025 (UTC)[reply]

And ANI didn't recognise it to be LLM writing anyway. --Gurkubondinn (talk) 23:18, 17 December 2025 (UTC)[reply]

Just one person at ANI, who does not appear to know what they're talking about -- the whole thing is one continuous AI tell, as are the other course descriptions by this professor. (also, like... "it could be the textbook manufacturer"? What textbook? This is a seminar, not Algebra 101. Black Feminist Thought isn't a textbook.) That said, it's not really our business to mess with professors' old syllabi. Gnomingstuff (talk) 02:01, 18 December 2025 (UTC)[reply]

Agree, this is out of scope. Professors using llms to generate course syllabi on obscure WikiEd pages is between them and WikiEd and their students. CMD (talk) 03:04, 18 December 2025 (UTC)[reply]

I apologize for taking it to WP:ANI. I still don't know how Wikipedia works in full. ~2025-31416-56 (talk) 12:53, 18 December 2025 (UTC)[reply]

That's okay, it's a learning process and people won't hold a misplaced post against you. CMD (talk) 12:58, 18 December 2025 (UTC)[reply]

"Delve into a rich tapestry"... good one. I guess nobody has taught these LLMs not to mix metaphors. Einsof (talk) 03:34, 18 December 2025 (UTC)[reply]

User:Feizan

Feizan (talk · contribs · deleted contribs · logs · filter log · block user · block log)

Many telltale signs of AI. Huge edits using curly quotes and awkward tone, while talk page responses are sloppy. Thank you! Magnolia677 (talk) 15:31, 18 December 2025 (UTC)[reply]

I had to deal with some rough edits to Porcupine but left the user alone due to inactivity past the end of November. More attention would be helpful now that they are editing more again. -- Recon rabbit 18:52, 18 December 2025 (UTC)[reply]

Regarding the Automatic Model Router for ChatGPT

Hey, guys. For your information, SuperPianoMan9167 has made an announcement on the AI Cleanup talk page that OpenAI has ditched the automatic model router for ChatGPT users on its free and $5 plans. Thoughts on what should be done here? sjones23 (talk - contributions) 00:21, 19 December 2025 (UTC)[reply]

Most (>60%) ChatGPT users will get GPT-5.2 instant (the fast model) by default, so it would not be unreasonable to see an increase in problems that are less pronounced in "reasoning" models. Other than that, not much changes. SuperPianoMan9167 (talk) 00:32, 19 December 2025 (UTC)[reply]

Rock Candy Mtn

Rock Candy Mtn (talk · contribs) is the latest newcomer to mass rewrite article leads using LLMs. I caught this when they copy and pasted raw LLM output into Chemolithoautotroph, which is on my watchlist after previously being targeted by now-CBAN-ed (from article space) Jaravedr in exactly the same way. Rock Candy confirmed [20] their use of LLMs when I asked them on their talk page. They said they would stop using LLMs but have continued to rapidly edit articles, including introducing a WP:V failure that I documented at [21]. I don't know if they are still using LLMs, although they very well might be given they falsely claimed [22] they were reviewing their mass-rewrites. The fantasy that PAGs requiring "review" of LLM-generated content will be sufficient to solve this problem lives on! All of their contributions require review; all of their contributions prior to [23] (yes, they found out about the lead maintenance tags at WP:TEAHOUSE) likely require reversion. NicheSports (talk) 03:37, 19 December 2025 (UTC)[reply]

newcomer tasks strikes again, who would have predicted gamifying edits would encourage this Gnomingstuff (talk) 04:30, 19 December 2025 (UTC)[reply]

Yikes! This is bigger than I realized. Thank you for looping me in. I think I'm out of my depth here.... time for me to step back and rethink. I'm glad you shared this with me. Rock Candy Mtn (talk) 16:33, 19 December 2025 (UTC)[reply]

I mentioned this at your talk page as well, but can't hurt to say it twice. This noticeboard is for cleanup, not for sanctions, and people here are not going to hold initial mistakes against you. I appreciate your collaborative approach at your talk page. You should stick around! Although please avoid LLMs :) They add WP:V violations at higher rates than can realistically be cleaned up, as you can see from how many posts there are here. NicheSports (talk) 16:58, 19 December 2025 (UTC)[reply]

@Magnolia677 I saw your thread below, but figured we could consolidate here NicheSports (talk) 16:43, 16 January 2026 (UTC)[reply]

I didn't see this. Thanks. Yes, huge edits which look AI. Numerous talk page warnings and denials. Magnolia677 (talk) 17:06, 16 January 2026 (UTC)[reply]

The LLM use is continuing, now coupled with repeated misrepresentation of that use.

Claimed usage: [24][25][26] it is not clear what this means but I am confident it is not honest
Actual usage: Create outline of draft [27], use LLMs to generate the entire article [28][29] (note the edit summaries), then make minor changes [30] if at all. Special:PageHistory/Inksters General Store (Winnipeg) is revealing.

Examples of what they're pushing to mainspace:

Their close proximity allows for comparative interpretation of domestic and commercial space within a single settlement property. sourced to this [31], completely fails verification
Its role within the settlement illustrates how small, multipurpose buildings supported both social exchange and economic resilience in early Red River communities during a period of gradual transition from rural river-lot farms toward more concentrated patterns of settlement. sourced to this [32], completely fails verification

I'm going to draftify all of their article creations that qualify, which I expect will be most. NicheSports (talk) 00:43, 17 January 2026 (UTC)[reply]

I've draftified Carberry Plains Museum, egregious model use, full of synth and unsupported statements alongside typical LLM prose.

@Rock Candy Mtn after stating you were out of [your] depth and that you would step back and rethink [33] things when this report was initially opened, you have continued to disruptively edit using a model. If you do not stop immediately you will likely end up blocked (or community banned as many persistent LLM editors have been lately), consider this a final warning before an ANI report is made: stop editing with an LLM, you have repeatedly demonstrated an inability to do so constructively. fifteen thousand two hundred twenty four (talk) 02:02, 17 January 2026 (UTC)[reply]

I draftified all of their remaining article creations after quickly finding similar V/SYNTH issues and many AISIGNS. NicheSports (talk) 05:06, 17 January 2026 (UTC)[reply]

Lots of WP:NEWLLM by ServiceAT

I stumbled across Mary C. Lawton while patrolling new pages. Aside from the obvious **markdown** and peacock words, they seem to have been creating articles very rapidly. A broader review of their contributions seems necessary, and I don't feel comfortable mass draftifying their creations myself. lp0 on fire () 18:19, 22 December 2025 (UTC)[reply]

Took a look -- the AI articles seem to start at least as far back as January 2024. It seems that some of them may be LLM translations. Gnomingstuff (talk) 21:07, 22 December 2025 (UTC)[reply]

I did not use any LLM in drafting WP pages prior to this week (December, 2025).

I am now going back to redraft my recent contributions "in my own words". I already did check references and facts and will do so again. I understand that I should have been aware of the prohibition on using LLMs to generate entire articles from scratch, but, with apologies, I was not. ServiceAT (talk) 21:29, 22 December 2025 (UTC)[reply]

Tracking subpage created, only edits from December were considered. 13 articles would benefit from a third party review. fifteen thousand two hundred twenty four (talk) 07:14, 23 December 2025 (UTC)[reply]

Apologies for the assumption. Were the earlier articles (~2024-ish) translations? Gnomingstuff (talk) 00:51, 24 December 2025 (UTC)[reply]

Phage therapy

Moved from Wikipedia talk:WikiProject AI Cleanup § Phage therapy

Hi, I am concerned about the phage therapy article and in particular (but not just) this edit. Checking the first few PMIDs in the citations added in that edit, they all point to different/unrelated articles so it appears that the references are made up. Thanks Rjwilmsi 08:55, 23 December 2025 (UTC)[reply]

Hello @Rjwilmsi, I've checked that and other edits by the same editor to phage therapy and there are numerous references with non-matching DOIs, PMIDs, and sometimes the article titles themselves don't exist (such as Innovations in Therapeutic Bacteriophage Development: From Bench to Bedside and Bacteriophage therapy: A potential solution for the antibiotic resistance crisis). These appear to be hallucinations. I've removed their additions to phage therapy in Special:Diff/1330173717.

Special:Diff/1194896302 also appeared to have possibly been LLM-assisted, but more likely is some form of COI. I've reverted it as well due to V, OR, and CLOP issues.

Nothing else in the edit history stands out to me personally. fifteen thousand two hundred twenty four (talk) 23:37, 29 December 2025 (UTC)[reply]

User:UrusHyby

UrusHyby (talk · contribs · deleted contribs · logs · filter log · block user · block log)

This editor first began adding AI-generated text to Litvinism with all the obvious signs. See for example this edit which is obviously AI-generated. Despite the warnings on their talk page, they have continued to make large additions to a range of articles with signs of AI use. It seems possible that they have now added AI-generated text along with fictitious references that do not directly support the statement. This is evidenced by some of the references having no page numbers or a long range.

See for example this edit to Battle of Orsha where they included a link to a page that does not exist (I could not find any such article that exists). After this was reverted, they restored the changes but added more sources that do not directly support what is being said. Similar issue at Marc Chagall. In this edit they added more sources and another editor tagged one of the statements with the following reason: Editor, who added this and two other refs, provided a non-existent quote in one of them, while twisting some other already present one in spite of quotation marks around it. Mellk (talk) 11:38, 23 December 2025 (UTC)[reply]

As another example, UrusHyby made this edit to Kievan Rus' and one of the statements they added was: It was a multi-tribal, loose federation where the Principality of Polotsk (in modern Belarus), the Principality of Kiev (Ukraine), and the Novgorod Republic (Russia) acted as key, often competing, centers of power. I looked at the cited source (they specified pages 10 to 15) and I could not find those specific states being mentioned as centers of power. For example, the Polotsk principality is only mentioned in the context of post-Soviet Belarusian intellectuals turning to it in "their search for the origins of their nation in the same historical period" (p. 12). Mellk (talk) 12:33, 23 December 2025 (UTC)[reply]

They've been using an LLM since they resumed editing in September, [34] is LLM nonsense. All references at their most recent creation, Mikita Melkazioraŭ, don't exist.

I've left them a message on their talk page that will hopefully convince them to desist. Cleanup is necessary. fifteen thousand two hundred twenty four (talk) 13:21, 23 December 2025 (UTC)[reply]

Hey everyone, thanks for your attention and your strive of having wikipedia the best source of corect information. That i share as well. I will copy here the comment i've made on my personal page, stating that I do believe that i follow the core concept of WP that AI is a proper tool, is it used carefully/. That is exactly what i was doing. Similarly the approach of WP:AICLEAN states "The purpose of this project is not to restrict or ban the use of AI in articles, but to verify that its output is acceptable and constructive, and to fix or remove it otherwise.". All the changes i do are goodwill based and supported by sources that i personally handpick.

But thank you for the notice I will be paying even more attention that the links remain intact after the automated proofreading\grammar corrections that i do. Preferences (talk) 14:08, 23 December 2025 (UTC)[reply]

All the changes i do are goodwill based and supported by sources that i personally handpick – All of the references you added to Mikita Melkazioraŭ are hallucinated and do not exist, this edit to Bicycle-sharing system added a hallucinated reference, as did this edit to ERM Telematics, as did this edit to Package tracking, as did this edit to Device tracking software, as did...

Nowhere in Wikipedia policy will you find it stated that AI is a proper tool, it is simply not a core concept. Two policies and guidelines which do exist are WP:VERIFIABILITY and WP:DISRUPT. Your editing is in clear continual violation of both. Your automated proofreading\grammar corrections are simply disruptive, and if you continue to perform them I will open an WP:ANI report myself. Stop.

If you lack the ability to proofread and correct grammar yourself, and instead must rely on copying and pasting output from an LLM, then you should not be making those edits at all. fifteen thousand two hundred twenty four (talk) 14:48, 23 December 2025 (UTC)[reply]

Even though, I am still confident about the sources I use. In all of the edits you've mentioned, there are proper links 1 for the Bicycle-sharing system, or this in Device tracking software edit I did det your point, as you're right, that some of the links get broken, unfortunately.

Which is a pity, as many of the articles that I improved were in a really poor shape, including a politically-hot one Litvinism that was a subject for deletion. And having good intentions and some understanding of subjects I tried to improve it as well as some others.

I do doubt that your revert of my edit of ERM Telematics article that had WP:PROMO since 2016 made any good, and brought it any to a better state, comparing with my version, even keeping in mind usage of AI tools.

But, I got the point. Thanks again. I will be more careful. Preferences (talk) 15:19, 23 December 2025 (UTC)[reply]

Are you sure you are confident about the sources you use? What about this edit, where you provided a non-existent quote "Born Moishe Shagal in Liozna, near Vitebsk, in what is now Belarus" - allegedly from the Jackie Wullschlager's book? There is no such a quote in the book! So where did you find that quote? Alexschneider250 (talk) 20:25, 23 December 2025 (UTC)[reply]

In the very same edit you added a ref to the book "Minsk - Wilna - Stationen einer undenkbaren Freundschaft", allegedly by Thomas M. Bohn. Are you sure that book exists? If so, could you provide a link to a web page saying anything about that book? Alexschneider250 (talk) 22:43, 28 December 2025 (UTC)[reply]

As one more example, UrusHyby made this edit to Russification adding a quote (with quotation marks around it): "What the Russian bayonet didn't accomplish, the Russian school will.", allegedly by the Russian Governor-General Mikhail Muravyov, from the book "Belarus: A Perpetual Borderland" by Andrew Savchenko, p. 55. But I read that entire page (and neighboring pages too) in that book and I could not find any sentence like that there, not even slightly close in meaning to that quote.

Moreover, that quote is a fake! Mikhail Muravyov never said that. Actually, Ivan Petrovich Kornilov, a Russian official in the 19th century, said this:

Русское образованiе сильнѣе русскаго штыка. Въ какiе-нибудь 4 года русскiя школы сдѣлали болѣе для образованiя народа и ослабленiя полонизма, чѣм войска въ десятки лѣтъ. [Russian education is stronger than the Russian bayonet. In about 4 years Russian schools did more for the people's education and weakening of the polonization than troops in tens of years.]

— Ivan Petrovich Kornilov, Kornilov, I. P.: Russian Cause in the Northwestern Krai (St. Petersburg, 1901), A.P. Lopukhin's Press, p. 147.

This quote is being twisted and spread all around Wikipedia by politically-engaged editors, who are also trying to make it look like it was said about the oppression of the Belarusian language. Obviously, the quote is about using education to oppose the imposition of the Polish language among inhabitants of the Northwestern Krai. Alexschneider250 (talk) 00:56, 29 December 2025 (UTC)[reply]

Tracking subpage created, 92 articles appear to warrant further review. While compiling the list, I encountered a multitude of hallucinated references. fifteen thousand two hundred twenty four (talk) 18:35, 23 December 2025 (UTC)[reply]

This is continuing, see Special:Diff/1333034795. Most of this does not verify in the source [35], and whatever does is SYNTH. Very likely LLM-generated. NicheSports (talk) 16:38, 16 January 2026 (UTC)[reply]

They have now reinstated this content [36]. I left them a final warning on their talk page [37] NicheSports (talk) 09:56, 20 January 2026 (UTC)[reply]

Talk:Arthur Katalayi

Large amounts of likely AI-generated uncivil talk page messages left on this page. I left a comment under the first one, but I decided not to further on, since we shouldn't feed the trolls. ~2025-31416-56 (talk) 13:20, 23 December 2025 (UTC)[reply]

AuroraGlyph 92

AuroraGlyph 92 has been making AI-generated edits (utm_source=chatgpt.com in links here and subsequently here, which I reverted and warned about because they read like advertisements to me). I only found two edits with the utm_source (though I didn't check all of their edits), but some of their other edits feel like it, like this one, which also has a source with an access date in 2020, and possibly this one. Not all of the edits they have made seem AI-generated to me? Sorry if this isn't enough to warrant being posted here. Crow Basket (talk) 22:31, 23 December 2025 (UTC)[reply]

Looks like possibly another LosAngeles000! sock, same area of articles and pattern of adding AI-captioned images Gnomingstuff (talk) 00:42, 24 December 2025 (UTC)[reply]

Editor has now been blocked as a sock, as their edits were made in violation of a block or ban, they may be WP:BANREVERTed. fifteen thousand two hundred twenty four (talk) 06:37, 24 December 2025 (UTC)[reply]

Lehrling18 creating AI-generated pages

Hi all,

The user I mentioned above has been creating drafts and revising existing articles (380 edits, account created 22 November) about French WW2 military units (typically French Army corps). Interestingly, he has also participated in talk pages ([38], [39], [40]), with AI-generated responses.. I haven't done a deep dive into whether his sources are hallucinated, but none of them have URLs or are CS1. He has also been adding navboxes to the articles, but they are made from tables + CSS (they won't uniformly update) and occasionally are named differently depending on the article.

Here are some of the articles that he has created. While it isn't as obvious compared to other AI-generated pages, the signs are still there and his talk page participation is damning. In draftspace, when his content gets denied for being AI-generated, is appears that he just regenerates it then resubmits it.

EatingCarBatteries _{(contributions, talk)} 21:07, 24 December 2025 (UTC)[reply]

This LLM-generated text has been collapsed and should be excluded from assessments of consensus.

The following discussion has been closed. Please do not modify it.

I'm the user in question. I'm a programmer, which explains my workflow: I use structural skeletons and batch-processing to handle large datasets efficiently. It's not AI.

Regarding the feedback:

Templates & Collapse: I've already started replacing the CSS tables with proper templates and removed the "click to expand" feature to meet MoS standards.
Sources: I use physical archives and specialized literature for 1940 OOBs. They don't have URLs, but I'll use CS1 formatting going forward.
Experience: I've done extensive work in other Wikis before; I'm not a bot.

Note on timing: I received a message on my talk page at 20:41 (UTC) on Dec 24. I was reported here at 21:07 (UTC) the same day—only 26 minutes later, on Christmas Eve. I would have appreciated more than half an hour to respond to technical feedback before being publicly accused of using AI. I'm currently building a template library on my user page to fix all remaining issues. Lehrling18 (talk) 01:07, 25 December 2025 (UTC)[reply]

It is generally a bad idea to respond to allegations of AI use with AI.

I can tell because of the bolded list formatting. SuperPianoMan9167 (talk) 01:19, 25 December 2025 (UTC)[reply]

It's not AI. – You are using an LLM, if you continue to lie and edit in this manner using an LLM, an ANI report will be opened which will most likely result in you being indefinitely blocked or cbanned. Stop. fifteen thousand two hundred twenty four (talk) 06:57, 25 December 2025 (UTC)[reply]

Weirdly, they seem to have vanished immediately after they were reported here. Prior to this, they were making multiple edits per day. ~2025-31416-56 (talk) 19:05, 9 January 2026 (UTC)[reply]

Their userpage also appears totally AI generated. Sarsenet•he/they•(talk) 09:08, 25 December 2025 (UTC)[reply]

Bank Pages using LLM

Originally posted here but was advised this would be a better location. A lot of pages linking from Regional rural bank have been created and/or heavily edited recently in what I believe are using LLMs. I verified through online AI checkers which confirm prior to moving a few to draft, but one of the users stated they are not and moved them back to main space. Hoping for more opinions as they are pretty bad (unsourced, promotional, etc.). See Haryana Gramin Bank, Chhattisgarh Gramin Bank, Jammu and Kashmir Grameen Bank, Rajasthan Gramin Bank, and Gujarat Gramin Bank. A couple are at AfD but if they are kept will need cleanup. Posting here since the creator removed the AI notice from one of the pages.--CNMall41 (talk) 01:59, 26 December 2025 (UTC)[reply]

There are two users involved here: Ankit231132 (talk · contribs) and Ajitsinghbhagat (talk · contribs). The users are reverting each other [43] so I don't suspect off-wiki coordination. They are both editing disruptively but seem to be doing so in different ways.

Ankit231132 (talk · contribs). I don't know if this editor is using LLMs but they are definitely adding copyvio. The first version of Chhattisgarh Gramin Bank confirmed violations when I ran it through earwig [44]. The first version of Jammu and Kashmir Grameen Bank doesn't look LLM-generated (except for the Logo section, that was either LLM or - more likely - copied from something earwig couldn't find). This comment on their talk page [45] is also not LLM generated. Special:Diff/1328503718 might be a hybrid of LLM/human written though. But unless someone can dig up some diffs showing clear LLM misuse in article space I would treat this as a copyvio case. I don't know if it is serious enough for a CCI request, my suggestion would be to look for a few more clear violations and then drop a note on MCE89's talk page and ask what to do about it.
Ajitsinghbhagat (talk · contribs) is definitely misusing LLMs, which I have warned them about 4+ times at their talk page and Talk:Disney Star - see Talk:Disney Star § Issues with recent probable LLM-generated edits for several documented examples. This user hasn't created as many bank articles as Ankit (their focus seems to be adding LLM slop to Disney Star), but the ones they have, such as Draft:Meghalaya Rural Bank, show strong signs of LLM content. The difference with Ankit's creations is pretty clear - Ajit's are in perfectly grammatical English, Ankit's are not.

NicheSports (talk) 05:07, 26 December 2025 (UTC)[reply]

Hi @NicheSports, thanks for the ping. I started going through Ankit231132's contributions to see if there was enough for a CCI request and ended up just doing most of the clean up in my sandbox (User:MCE89/cci) since there wasn't too much to go through. I've still got a couple of pages to look at, but as you can see there was definitely a lot of copyvio. To keep things simple I've only removed text that I could confirm as copyvio for now, but will go through and remove the rest of the unsourced/suspicious text. I agree that I didn't see too much evidence of clear LLM usage, although I'm not sure either whether the logo section they've added to most of these articles is copyvio or LLM-generated (or copyvio of an LLM-generated source?). MCE89 (talk) 02:36, 27 December 2025 (UTC)[reply]

I don't think I pinged you but I suppose I did light the bat signal :) Thank you for answering (and more!), this is incredibly helpful. The logo section is a question but my intuition says this was copied, not generated by this user. It doesn't really matter because it is unsourced promo and can be removed indiscriminately. @CNMall41 looks like MCE89 has handled Ankit's edits, I can help you with Ajit's tomorrow. NicheSports (talk) 03:25, 27 December 2025 (UTC)[reply]

Glad to help! and linking to someone’s user page in a signed post does send a ping, just FYI :) MCE89 (talk) 04:19, 27 December 2025 (UTC)[reply]

Whoops. Not sure how it took me the better part of a year to figure that out. But thank you again! NicheSports (talk) 04:51, 27 December 2025 (UTC)[reply]

Appreciate the help from both of you. Once the AfDs are done, I will go back through and remove anything that is unsourced and do any additional cleanup that may need to be done. Outside of copyvio, the content screamed WP:AISIGNS and with the OP removing the tags wasn't sure how to handle outside of ANI.--CNMall41 (talk) 06:52, 27 December 2025 (UTC)[reply]

Ajitsinghbhagat is being discussed at WP:ANI#Repeated disruptive editing. fifteen thousand two hundred twenty four (talk) 13:24, 27 December 2025 (UTC)[reply]

I cleaned up Ajitsinghbhagat's LLM-generated mainspace edits. Also, I confirmed that part of the Logo section these users have been adding to articles is copyvio. See the pop up when you first go to [46] NicheSports (talk) 22:44, 27 December 2025 (UTC)[reply]

I would like to stubify them all based on the unsourced content and the fact two edits would rather bludgeon (removal of LLM tags, moving back to mainspace, etc.) instead of doing the cleanup of their mess. However, I took two of them to AfD so do not want to be accused of bludgeoning myself. I did a few such as Arunachal Pradesh Rural Bank which I took from this to this. And, tagged here for West Bengal Gramin Bank as you can see the clear LLM (bolded words and use of already linked topics in the see also section - Also a lot is unsourced so likely LLM generated). Also took Puducherry Grama Bank from this to this. Dozens more need addressed. That is, assuming notability is established for all. --CNMall41 (talk) 20:50, 26 December 2025 (UTC)[reply]

User:Pritik Wiki

Found the named user at the Teahouse asking a question about the now-blanked draft at User:Pritik Wiki/sandbox which was obviously AI generated, along with their responses. I also see warnings on their talk page about other edits being AI generated. I'm going to look through some of their recent edits to see if they need reversion, but if the tracking subpage wizards could make a tracking page that would be great. Athanelar (talk) 11:10, 26 December 2025 (UTC)[reply]

On investigation, much of their contribs prior to gaining EC seem to be minor copyedits (likely gaming, which I've already reported at ANI) so the scope of cleanup here might not be very high at all. Athanelar (talk) 11:24, 26 December 2025 (UTC)[reply]

Courtesy link: WP:ANI § EC gaming by User:Pritik Wiki

Their three non-gaming articlespace edits have already been reverted, and their gaming edits don't appear to involve LLM use. Marking cleanup as unnecessary at this time, but leaving the report open in case they make more edits in the near future. fifteen thousand two hundred twenty four (talk) 13:28, 26 December 2025 (UTC)[reply]

The copyedits are in a way more frustrating than full articles. I'm going through some rapidfire masses of AI copyedits currently, and it's such a mixed bag; I can't really bring myself to care about stuff like this even though it's almost certainly AI based on the edit history, but then you also have stuff like this which is obviously problematic (but already fixed). Or stuff like this, which does absolutely none of what the edit summary claims it does but is ultimately kinda whatever, next to this, which inserts promotional language while claiming to remove it. Gnomingstuff (talk) 22:45, 31 December 2025 (UTC)[reply]

I agree. Most of Pritik's copyediting was just inserting a punctuation mark or two, but I have seen AI copyedits merge two sentence clauses together in a way that changed the meaning entirely; I can't remember exactly what it was now, but it was the same effect as changing "Let's eat, grandma!" to "Let's eat grandma!" Athanelar (talk) 22:50, 31 December 2025 (UTC)[reply]

Hallucinating sources

Started finding what appear to be hallucinated citations – here, here, and here. (Even if the individual titles correspond to actual books, other details like authors and ISBNs did not match at all.)

Flagged to the editor who is experimenting with LLMs. They have now reverted a bunch of their own edits but think this series of articles is worth a closer look to make sure there aren't any other garbled or made-up citations. Cielquiparle (talk) 13:03, 30 December 2025 (UTC)[reply]

Tracking subpage created and is available at the top of this report. 158 pages need review. I've done my best to exclude obviously unproblematic edits. fifteen thousand two hundred twenty four (talk) 13:44, 30 December 2025 (UTC)[reply]

I've given them a bespoke final warning, if they misuse an LLM again please ping me. fifteen thousand two hundred twenty four (talk) 14:10, 30 December 2025 (UTC)[reply]

@Fifteen thousand two hundred twenty four They continue to edit without checking their own work or taking responsibility per their Talk page. Seems like they are WP:NOTHERE to build an encyclopedia. Cielquiparle (talk) 19:26, 25 January 2026 (UTC)[reply]

WikiNoia is now indefinitely blocked from articlespace following an ANI report at WP:ANI#Another continually unconstructive LLM editor. fifteen thousand two hundred twenty four (talk) 01:02, 26 January 2026 (UTC)[reply]

There exists unanimous consensus to allow for presumptive reverts of all WikiNoia's past articlespace, non-minor edits, see WP:ANI#Presumptive revert proposal. When making a presumptive revert, I would suggest leaving a link to that consensus in the edit summary. fifteen thousand two hundred twenty four (talk) 23:43, 26 January 2026 (UTC)[reply]

Flemza

Flemza (talk · contribs) seems to be using LLMs with minimal subsequent review to create questionably notable [47] articles on individual songs. I just draftified everything I could but many others were created from redirects and aren't eligible. The user acknowledged LLM use at User talk:Flemza § LLM use and claims they were reviewing the outputs [48], but I am unconvinced by their responses [49][50] at Draft talk:No Apologies (Papa Roach song) § LLM generated content where I documented several examples of problematic and likely LLM-generated content. They also moved an article back to mainspace after a rapid rewrite [51] that introduced a material copyright violation. They have not answered my question [52] about whether this violation was caused by LLM or by them. NicheSports (talk) 20:53, 31 December 2025 (UTC)[reply]

Sascho Jovanoski 70329

Sascho Jovanoski 70329 (talk · contribs · deleted contribs · logs · filter log · block user · block log) has made hundreds of edits, almost all of which seem to be LLM-generated. These include creating pages (such as Draft:IFRS 19, which I draftified). These edits seem to be largely unreviewed, as evidenced at IAS 37#2. Discounting and Unwinding (Time Value of Money), saying [Image illustrating the unwinding of a discount for a provision over time under IAS 37]. While the warnings left are recent, these edits are very large and there are a lot of them, which is why I'm bringing this here. Also pinging @Iseult who originally found this. HurricaneZeta^C 22:01, 1 January 2026 (UTC)[reply]

The edit summaries seem like chatbot responses, for instance this:

Certainly. Here is the English summary of the strategic pillars of the EU's financial market policy for a Wikipedia article, based on your categorized list. 🇪🇺 Strategic Pillars of EU Financial Market Policy The European Union's financial market policy, primarily executed by the Directorate-General for Financial Stability, Financial Services and Capital Markets Union (DG FISMA), is structured around four main strategic pillars. These pillars aim to enhance EU competitiveness, ensure financial s

Gnomingstuff (talk) 22:33, 1 January 2026 (UTC)[reply]

They were warned, at great length, in December that their edits were problematic, and that AI/LLM editing was forbidden and that their frequent and large editing was disruptive. They then removed all these discussions without a satisfactory response to the AI/LLM usage, and I see they've continued without any change to their fast and large editing behavior, and their talk page has continued with notices of issues. Disappointing. Zinnober9 (talk) 22:33, 1 January 2026 (UTC)[reply]

Good catch. I think that ANI is defensible and more than so if editing continues. Iseult^Δxtalk to me 01:02, 2 January 2026 (UTC)[reply]

I did sort of get this editor to admit to using an LLM. Even asked them to clean up after themselves, but of course never got an answer. --Gurkubondinn (talk) 19:18, 2 January 2026 (UTC)[reply]

They also posted § Feedback request for manual draft in my sandbox on my talk page, but ignored me when I replied. --Gurkubondinn (talk) 19:26, 2 January 2026 (UTC)[reply]

Tracking subpage created which is available at the top of this report. 98 pages appear to need further review. fifteen thousand two hundred twenty four (talk) 22:58, 1 January 2026 (UTC)[reply]
They've created a number of mainspace pages themselves with these issues. Should these be draftified or taken to AfD? HurricaneZeta^C 00:30, 2 January 2026 (UTC)[reply]
If the problematic page was created directly in mainspace, then draftifying, tagging, and leaving a notice on the talk page would be my suggestion. Any pages which are WP:G15 eligible should be tagged in place. fifteen thousand two hundred twenty four (talk) 01:05, 2 January 2026 (UTC)[reply]
I think I got all the G15s, working on draftifying the ones they created. One went through AfC and seems fine, so I just tagged it. HurricaneZeta^C 17:47, 2 January 2026 (UTC)[reply]

Courtesy ping to @Dormskirk regarding this thread as they are continuing LLM edits such as [53]. HurricaneZeta^C 16:34, 10 January 2026 (UTC)[reply]

Pblocked as of the 10th. Iseult^Δxtalk to me 22:26, 12 January 2026 (UTC)[reply]

Likely AI translations by User:Leeanah with some issues

The edit summary for this initial version of Transport in New Caledonia claims it's a translation from the French Wikipedia article. I don't speak fluent French... but you don't have to speak French to notice that in the French article, at time of creation, the "Transport fluvial" section was empty besides a template to expand it. That section of the English article has a lot of text there, and it sure reads like AI text.

I don't know whether the LLM is doing this unprompted or if it was prompted to fix the "expand" tag -- the rest of the article isn't that obvious as AI, but I don't know French well enough to say how faithful it is. Either way though that's not great, and they have several other translations. Someone asked previously on their talk page a while back about whether they used AI, but they didn't answer. Gnomingstuff (talk) 18:57, 2 January 2026 (UTC)[reply]

When creating Correction of the Rhône upstream of Lake Geneva they added a utm_source=chatgpt.com ref which was absent from the corresponding frwiki article. Same with History of Suresnes. They've used chatgpt outside of translation tasks as well. I don't speak French either, but it is clear that the rich in preserve a rich and varied folklore from Acadian folklore isn't in the original frwiki text (and neither is the *relevailles* markdown). Not looking good.

This is the second time an OKA member has been the subject of a noticeboard post in recent days due to possible translation issues and lacking communication, see this ANI report. In that report the OKA founder, 7804j, was very responsive, so pinging them may be an option if Leeanah doesn't engage, and Leeanah really needs to engage. fifteen thousand two hundred twenty four (talk) 19:29, 2 January 2026 (UTC)[reply]

Thanks for the info. (to be clear I stumbled across this article by searching for keywords, not seeking out anything from that project specifically) Gnomingstuff (talk) 19:35, 2 January 2026 (UTC)[reply]

It seems too faithful in parts. For example, translating "route transversale" very literally into "transverse road". No idea if that is a manual translation issue or a machine translation issue, I have seen it occur in both. The presence of "<---- second part ---->" is quite odd though, and does suggest multiple chunks were translated separately and stuck together. CMD (talk) 10:19, 3 January 2026 (UTC)[reply]

Thank you for taking the time to look into this and for laying out the concerns in detail.

This edit was made back in August 2025, and I want to be clear that the “River transport” section should not have been added the way it was. At the time, I misunderstood the boundaries of a translation task and treated an empty section in the source article as something that could be developed further. That was an error on my part.

I did consult external sources then, but after reading the discussion here, it’s clear that parts of the content were not properly supported by those sources and should not have been included. Regardless of my intent, the end result does not meet Wikipedia’s standards for verifiability or for faithful translation.

As a first step, I will remove the “River transport” section entirely, as well as any other content identified as unsupported, so that the article reflects only what is present and sourced in the French version.

I’ve also changed how I work since then. I no longer expand or supplement content during translation. My role is strictly limited to translating existing, sourced text, with AI tools used only to assist with revising language and editing formatting, followed by manual review against the original article.

I also recognize that I should have responded sooner on the talk page, for that I apologize. While I’m a professional translator, I’m not the most experienced when it comes to navigating talk pages and Wikipedia’s internal processes more broadly. That’s something I’m actively working to improve, so I can engage more effectively and avoid situations like this in the future.

I appreciate the feedback from everyone here and take responsibility for fixing this and making sure it doesn’t happen again. Leeanah (talk) 15:59, 11 January 2026 (UTC)[reply]

Is it valid to use "AI tools used only to assist with revising language and editing formatting, followed by manual review against the original article"? I suppose the answer depends on how much "manual review" is really done. (What is "editing formatting"?) David10244 (talk) 03:52, 18 January 2026 (UTC)[reply]

with AI tools used only to assist with revising language and editing formatting, followed by manual review against the original article. This is evidently untrue, as evidenced by the fact that there are diffs above showing you added utm_source=chatGPT parameters, which only occur when source links are copypasted directly from ChatGPT. Either ChatGPT inserted a source into its 'edited formatting' which you copypasted without realising you were doing so, or you are not being entirely truthful about the scope of your AI usage.

It is also worth noting that your response here reads as at least partially AI generated in itself. In particular, language like As a first step... and the insistence that AI was only used for copyediting are both hallmarks of AI-generated responses.

@7804j I'm having serious concerns about your standards for the editors you employ via OKA. If you want your project to be taken seriously as a constructive part of Wikipedia, this kind of conduct by one of your translators is hugely problematic. Athanelar (talk) 12:56, 24 January 2026 (UTC)[reply]

Proposed deletion of History of U.S. involvement in Venezuela's petroleum industry

This page was seemingly created with an LLM (has ?utm_source=chatgpt strings and self-citations to other Wikipedia articles), but has since been edited by humans. I proposed it for deletion since it falls afoul of WP:NEWLLM, but if someone wants to take a crack at cleaning it up, please go ahead. There should likely be an article on this subject, just not contaminated with LLM slop. Einsof (talk) 23:39, 3 January 2026 (UTC)[reply]

If it has been substantially edited by humans then it does not violate NEWLLM. SuperPianoMan9167 (talk) 00:07, 4 January 2026 (UTC)[reply]

The instruction from NEWLLM reads, in its entirety, "Large language models should not be used to generate new Wikipedia articles from scratch". It makes no exception for what happens afterward. Einsof (talk) 00:10, 4 January 2026 (UTC)[reply]

Seems better suited for AfD than for prod, since this is very much not an uncontroversial topic nor of uncontroversial notability Gnomingstuff (talk) 00:32, 4 January 2026 (UTC)[reply]

Quite possible. I figured that since there are so many eyes on this topic right now, there is plenty of opportunity over the next week for anyone who objects to PROD to convert it to AfD. Einsof (talk) 00:39, 4 January 2026 (UTC)[reply]

I agree that there should be an article and will be trying to my best to rewrite it over the course of the week! I've rewritten the lead and the first section already.

Seems like there's already been a few edits to remove all the Wikipedia sources, although many unverified claims remain. I'm hoping to replace almost all of the prose in my changes. Altoids0 (talk) 22:05, 4 January 2026 (UTC)[reply]

Communitylover

Communitylover (talk · contribs · deleted contribs · logs · filter log · block user · block log) says they are a teacher working with their class to create an article on a local business. The draft they created tripped 1346 (hist · log) for a URL with a chatgpt parameter and I think is mostly raw LLM output - see Draft:Hyde Park Hair Salon Barbershop of Hyde Park. Over-attribution everywhere. I asked them about it on their talk page and they told me, in a message [54] I am about 98% sure was LLM-generated (gptzero says 100%), that they had not used an LLM. They then doubled down on this in their own words [55] - note the difference in writing style. I've never filed here for an editor based on a single draft, but this user seems sufficiently non-collaborative that I'd like a second opinion and (potentially) some help managing the situation. NicheSports (talk) 04:58, 4 January 2026 (UTC)[reply]

I've left them a comment also. They've been provided with relevant guidance, I'd wait and see what they do with it. fifteen thousand two hundred twenty four (talk) 05:44, 4 January 2026 (UTC)[reply]

Thanks fifteen. I saw your message on their talk page - very helpful NicheSports (talk) 05:46, 4 January 2026 (UTC)[reply]

User:CZmarlin and promotional edits

This user has made, and continues to make, a lot of articles and substantial rewrites of articles, mostly regarding cars and colleges. Going back to at least early 2025, they appear to be AI. They have denied it, sort of (although without directly denying that they used AI, as opposed to the books/articles cited) but I don't believe this, based on the content at hand.

Most of their substantive edits claim to be "CE" (copyedit I assume), but they are not just "copyedits" as they add substantive content. That content generally shows the bog standard WP:AISIGNS. A sampling:

Rambler Rebel: like tacking promotional participials onto sentences (, offering a compelling alternative to consumers., , showcasing AMC's technological initiatives., etc), adding editorializing not present in the original text, e.g., turning "Development of AMC's new overhead-valve V8 engine began..." into A crucial component of this new direction was the development... (also note crucial) I assume it has been hand edited somewhat -- how else does one get This change marked a pivotal for AMC -- but still.
Siedlce University: Promotional AI stuff all over the place, the most glaring being Its history, from its origins as a teacher-training college to its current status as a comprehensive university, underscores its enduring dedication to education and its impact on the Siedlce region and beyond.
Concord University: Undue emphasis on symbolism (its journey reflects the broader development, superficial analyses (reflecting its evolving mission, underscored the state's growing commitment to public education, signaling a new chapter in its long-standing commitment...), promotional tone and AI vocab all over the place, etc.
AMC Matador: The same stuff: The new Premier represented a new direction for the former AMC, showcasing a more modern, European-influenced design and engineering, reflecting the evolving automotive landscape of the late 1980s..

I haven't left a new comment because based on the above response I do not expect it to go well. Gnomingstuff (talk) 16:43, 5 January 2026 (UTC)[reply]

AMC Hornet: The AMC Hornet emerged as a pivotal vehicle in AMC's strategy, directly replacing the venerable ..., However, its robust and versatile rear-wheel-drive platform continued to serve as ... Furthermore, demonstrating the enduring adaptability of its original design, the Hornet's chassis underpinned the innovative ...
Nash 600 The 1949 Nash 600, with its daring aerodynamic styling, spacious interior, and innovative comfort features, represented a bold step for Nash Motors in the post-war automotive landscape, establishing a distinct identity in the competitive economy car segment, This innovative placement aimed to bring essential gauges closer to the driver's line of sight, a concept ahead of its time ... Nash also introduced a groundbreaking "Twin Bed" feature ...
Jerrari: These unique hybrids combine the rugged utility of a Jeep Wagoneer with the exotic power of a Ferrari, The 1977 Jerrari Wagoneer adopted an understated approach ... Modifications to the exterior were subtle yet functional, The 1977 Jerrari stands as a testament to Harrah's unique vision and ... where it continues to captivate with its blend of utilitarian ruggedness and supercar heart.

They've added a fair amount of unencyclopedic prose infused with common LLM-isms. I'd be curious to hear about their process in crafting the above samples. fifteen thousand two hundred twenty four (talk) 20:06, 5 January 2026 (UTC)[reply]

Gnomingstuff & Fifteen thousand two hundred twenty four, any further thoughts on this one? The edits are ongoing. Thebiguglyalien (talk) 🛸 01:44, 19 January 2026 (UTC)[reply]

I've notified them of this discussion, hopefully they can enlighten us a bit as to what their process was when creating the above text. The prose is less than encyclopedic, LLM or not. fifteen thousand two hundred twenty four (talk) 03:31, 20 January 2026 (UTC)[reply]

apologies, have been working on other backlogs, looks like not much to do at this point in time Gnomingstuff (talk) 20:29, 22 January 2026 (UTC)[reply]

Preye douglas

Preye douglas (talk · contribs) is a non-native English speaker [56][57][58] who is seemingly using LLMs for all of their major mainspace edits, including many article creations [59][60][61][62], as well as using LLM for comments on talk pages [63]. They have been warned about this multiple times on their talk page. They have wikilawyered about their use of LLMs [64]. I've been following this for a month but both the LLM use [65] and wikilawyering about it [66] (bonus, also LLM-generated) are continuing, so here we are. A few examples:

[67] The year began with Ruger's "Asiwaju" holding the top position for four consecutive weeks in January. Early 2023 saw collaborative success, with "Gwagwalada" by Bnxn, Kizz Daniel, and Seyi Vibez spending four weeks at number one, marking one of the year's most successful multi-artist collaborations.: The editorializing about collaborative success is unsupported by the source [68]
[69] Together, these metrics reflect the overall consumption and popularity of music within Nigeria's contemporary music landscape. has no in-line citation at all
[70] In accompanying press materials, the track was introduced as a mid-tempo afropop song that combines affirmational lyrics with a reflective tone I'll let you guess whether the bolded bit verifies in the source [71]

NicheSports (talk) 03:13, 6 January 2026 (UTC)[reply]

Gwagwalada (song) is considered a collaborative success as the song charted on multiple charts. It reached number one in Nigeria and it's certified 2x platinum. If that doesn't sound like success to you I don't know what would. Not many collaborations achieved that in 2023. As for the Turntable chart methodology there is a source for that added right after "radio and television plays across Nigeria."

The line from Commas was basically paraphrasing of what was said in the sources.

Over a gentle rhythmic beat with swelling strings; ‘Commas’ exudes breezy confidence with a hint of reflection from the 21-year old who, in just three years, is already growing to become one of the most vital African artists of her generation. - Music-News

“Commas” offers fans an intimate look into the vocal growth and journey Starr has embarked on over the last few years. Staying true to her positive and affirming lyricism, Starr’s latest single spotlights the singer’s continued focus on her growth as she sings, “I carry God so I fear nothing, Steady increasing the commas” on the single’s chorus. – OkayAfrica Preye douglas (talk) 03:47, 6 January 2026 (UTC)[reply]

This is pure WP:OR
The lead had no citations at all, and given there was no body in the article (only a table) all claims in the lead lack in line citation support
You did not provide the OkayAfrica article as a source for the third claim. Your statement above misrepresents the state of the article, whether you intended to or not. Please do not do that again

Did you use LLMs to assist with any of the edits I linked to in my post above? NicheSports (talk) 04:33, 6 January 2026 (UTC)[reply]

@NicheSports It's not WP:OR because "hint of reflection + affirmative lyricism" can be paraphrased as "affirmational lyrics with a reflective tone." Putting it that way, you are basically still saying the same without copying and pasting.

Typically, lead sections don't really require citations per WP:CITELEAD and that's just a list article. Apart from the chart methodology, whatever you see on there are facts derived from the table. The table has sources from the TurnTable magazine website.

The OkayAfrica source was cited in other parts of the article. I may not have added it as an inline citation after that line but hey, it's among the references on the page. Since you have a problem with that, you should have just added ^{[citation needed]} template after that line. But anyways, I will be adding that OkayAfrica reference now.

There is no denying that Artificial Intelligence has become a part of our daily lives, as mainly assistants. So do I use artificial intelligence in my daily life and work? Yes. Do I sheepishly follow whatever AI says without verifying if they are actually credible? No. Do I use LLMs in creating Wikipedia articles? Not so much just basically supportive tasks such as copyediting, and improving grammar but of course with caution as I am well aware of the tendency of these bots hallucinating at times and spreading misinformation.

Are those lines you listed written using LLMs? No. Those are basically the kind of lines you would find on any Wikipedia article about a record chart or song. Long before LLMs editors have always drawn inspiration from existing Wikipedia articles. Preye douglas (talk) 05:30, 6 January 2026 (UTC)[reply]

My comment about WP:OR was in relation to the first example in my original post, namely the phrases "Early 2023 saw collaborative success" and "marking one of the year's most successful multi-artist collaborations" NicheSports (talk) 05:51, 6 January 2026 (UTC)[reply]

That's basic English. It isn't that hard, really. The song was released in February 2023, that's considered an early part of the year. Still Q1. Now this collaboration goes on to dominate airplay and television charts in a country of 230 million people, topping the national chart and even getting plays in the UK to extent that it made top 10 on one of the country's official chart and still went ahead to shine on a billboard chart while being certified 2x platinum and nominated for Best Collaboration at Africa's version of Grammy Awards. That's huge success. It don't need to be on Hot 100 to be called a success :) Preye douglas (talk) 06:10, 6 January 2026 (UTC)[reply]

Do not combine material from multiple sources to state or imply a conclusion not explicitly stated by any of the sources. Similarly, do not combine different parts of one source to state or imply a conclusion not explicitly stated by the source. (emphasis mine)

Have you read the no original research policy? fifteen thousand two hundred twenty four (talk) 15:05, 6 January 2026 (UTC)[reply]

I have edited out that part, leaving the main quote from the artist and I have read WP:OR Preye douglas (talk) 19:55, 6 January 2026 (UTC)[reply]

Note: I forgot to link to the articles for the three examples I provided above. I added those above, permalinked to their most recent version at the time of my initial post NicheSports (talk) 04:08, 6 January 2026 (UTC)[reply]
There you go. No Wikipedia editor is perfect. Everyone makes mistake sometimes. The important thing is admitting it and fixing it or correcting whatever that was wrong. I make sure to double check my articles before publishing. I use sandbox to create. I cannot guarantee that there won't be mistakes but as a community why don't we work together to call out these mistakes when we see them or try to fix them ourselves. Some of the issues like not citing sources properly are things you can point out on the article talk pages and tag the editor to it so they can fix it if you don't want to do it yourself. And these kinds of mistakes are human error, some are caused by the editor not having enough time or probably they have a busy schedule so they weren't able to add all references. These kind of things happen on Wikipedia not trying to say it's a good practice but just saying that calling me out and attributing it to LLM is just unfounded because prior to the evolution of LLMs or proliferation of Artificial Intelligence, Wikipedia has always had citation issues this is just human error, as a matter of fact LLMs I believe have the capability to also cite sources so if truly I was using LLMs I'd probably just program it or train it to write and include citations as well and I bet you they would do it better than most editors here. Preye douglas (talk) 04:35, 6 January 2026 (UTC)[reply]
Do you use tools such as Grammarly, Quillbot, or DeepL? Those are AI-based which is not always clear in their advertising. Gnomingstuff (talk) 20:09, 6 January 2026 (UTC)[reply]
No, I don't use any of those tools. Preye douglas (talk) 23:29, 6 January 2026 (UTC)[reply]

Going 2x platinum and reaching number one in a country with over 230 million people is success. Not to mention the song penetrated a UK official chart and US billboard chart. Very few Nigerian songs get to that level so that's success judging from where the artists are coming from. Preye douglas (talk) 04:16, 6 January 2026 (UTC)[reply]

Since when did it become a good practice to copy and paste exactly what sources say. Experienced Wikipedia editors know that first you have to read what's in the source and deliver it in your own words while not straying very far from what the sources are saying. Preye douglas (talk) 04:21, 6 January 2026 (UTC)[reply]

More LLM translations, more issues

This is another set of translations from the OKA project (which has disclosed LLM use), but I found it via an unrelated search.

There are dozens of these but I'm just going to focus on Caciquism. This is the French article at that time. There is AI text that does not have any counterpart in it, most of which are synthesis tacked on to paragraphs, such as:

This linguistic borrowing highlights the historical and cultural connections between these various groups.

* This made them a crucial link during the era. (see below)

This anecdote depicts the workings of caciquism and the seizure of power by the two dynastic parties.
Additionally, the formation of important mass parties, such as the CEDA, marked a pivotal moment in political history. (CEDA is mentioned and called an important party but not the "pivotal moment" stuff)

This seems to be most of the obvious ones; my knowledge of French isn't good enough to assess the faithfulness of the actually translated stuff well. (It seems like some direct quotes are being paraphrased and/or "translated" into AI-isms, which seem somewhat anachronistic given the period of history we're talking, but that could well be from the French quotes.) Gnomingstuff (talk) 21:48, 7 January 2026 (UTC)[reply]

I know the head of the project has come forward to disclose use, but it looks like it’s just the tip of the iceberg regarding how much AI this project contains. The instructions on meta direct people to translate using Grok, then use other LLMs to improve the language. I can’t imagine the result being anything more than nonsense soup. By their own admission, the bulk of the work is done via LLM. ExtantRotations (talk) 02:40, 9 January 2026 (UTC)[reply]

Luckily, I'm a native French speaker so I should be able to help! The lead section is an accurate translation with sentences being split for readability, with the only three actual changes being Dans l'historiographie, ainsi que dans la presse ou les milieux intellectuels de l'époque (In historiography, as well as in the press or intellectual circles of the era) becoming In historiography, journalism, and intellectual circles of the era (removing the implication that it was the press of that era), très influent essai (very influential essay) being translated to influential essay, and that same sentence moving around leading en raison du ... (due to the...) becoming ... popularized the term. So, pretty okay lead translation all things considered.

In the #Concept of "cacique" section, the very first paragraph adds changes in meaning. Besides the hallucinated sentence pointed out by @Gnomingstuff, it removes the fact that the word in other Western European languages is a borrowing of the Spanish one (while that could be something not verified by the source, there is no indication of that being the case). It also adds a specifically (in It referred specifically to...) which doesn't carry any meaning or equivalent in the French original. The second paragraph is a very accurate translation, and I don't have anything to say about it. In the third paragraph, I'm confused by "seigneur" not being translated – the word doesn't have any specifically French meaning as we're talking about Spanish society primarily, and translating it as "lord" (while keeping the Spanish original "señor") would've made more sense. What leaves me more puzzled is the choice to move half of the later quote outside of the quotation marks while leaving it closely paraphrased.

The fourth paragraph is where we start to see more inaccurate translations, especially in quotes: personne puissante, qui jouit d’une influence par la crainte qu’il inspire dans une localité (a powerful person who holds influence through the fear they instill in a locality) becomes a dominating individual who instills fear and holds influence in a locality, which, while approximate, obscures the link made in the original quote between fear and influence. We also see qui a plus de commandement et de pouvoir, et veut par sa superbe être craint et obéi des inférieurs (who wields more command and power, and wants through his splendor to be feared and obeyed by inferiors) become who wields more power and commands more respect by being feared and obeyed by those beneath them, which completely reverses the causal link. The fifth paragraph and the two associated quotes are comparatively quite accurate. The last paragraph of this first subsection is also okay, although we again see the issue of quotes being very closely paraphrased without being marked as quotes. The other two very short subsections (three sentences in total) are fine too.

In the beginning of the second section, This made them a crucial link during the era. is surprisingly not an AI-ism, but a translation of Il devient toutefois un maillon véritablement essentiel au cours de... that got moved at the end of the paragraph. I've only briefly scrolled through the rest of the article, but I'm especially concerned at two long paragraphs, marked as unsourced in the French original, being imported without sources and with the tags removed. Chaotic Enby (talk · contribs) 11:24, 9 January 2026 (UTC)[reply]

oh good catch on the latter, I was mostly going sentence by sentence. For some reason LLMs seem to really hate direct quotations when used for editing, they often replace them with paraphrases, so that part isn't surprising. Gnomingstuff (talk) 14:22, 9 January 2026 (UTC)[reply]

The more worrying part is that the direct quotations are often very closely paraphrased (if at all) while not being marked as quotes in the output. Chaotic Enby (talk · contribs) 14:33, 9 January 2026 (UTC)[reply]

User making only LLM-generated contributions, unresponsive to talk page warnings & messages

(transferred from Wikipedia:Administrators' noticeboard/Incidents#User making only LLM-generated contributions, unresponsive to talk page warnings & messages)

User:Mudit23june has been making only LLM-generated contributions to a variety of astronomy articles, such as 2025 MN45, C/2025 R3 (PanSTARRS), and 2025 PN7. This user does not respond to talk page warnings and messages at all and does not edit articles to correct mistakes and citation issues. I suspect a violation of WP:LLM and WP:NOTHERE given their complete lack of response and corrections. Nrco0e (talk • contribs) 21:35, 8 January 2026 (UTC)[reply]

I've updated the subpage formatting, pages remaining should be automatically tracked. There was an edit conflict which I tried to properly integrate, apologies if I've missed anything. fifteen thousand two hundred twenty four (talk) 00:20, 9 January 2026 (UTC)[reply]

Thank you! It looks a lot nicer and more organized. Nrco0e (talk • contribs) 00:33, 9 January 2026 (UTC)[reply]

Camoz87

Camoz87 (talk · contribs) I've been following this one for a few weeks. This user is a non-native English speaker [72][73][74] who is relying on machine translation (at a minimum) for their mainspace edits. They have added transvio to articles [75] and have been warned about LLM use 5+ times, by 3 editors, on their talk page. They have created many articles, which are often then draftified by other editors [76][77]. This machine-generated text is insufficiently reviewed, sometimes including references with broken URLs [78][79]. Here is an example of a material source-to-text integrity issue that I identified [80]. This is even worse: it sure looks like they ran a few hundred poems through an LLM and asked it to summarize them in English. The user shows no signs of changing their approach, continuing to make machine-assisted edits today. [81] NicheSports (talk) 16:04, 9 January 2026 (UTC)[reply]

There's got to be a quicker way to sort cases like these out rather than a lengthy ANI thread followed by a CBAN Kowal2701 (talk) 17:07, 9 January 2026 (UTC)[reply]

Agreed. Although this one doesn't merit a CBAN, rather a mainspace block. It could have gone to ANI at any point in the last week I've just had ANI LLM fatigue. NicheSports (talk) 17:20, 9 January 2026 (UTC)[reply]

As an it.wiki sysop please consider that this user has been indef banned on it.wiki and is constantly evading the block creating hundreds of articles with LLMs even in his native language. --Friniate ✉ 20:18, 12 January 2026 (UTC)[reply]

Thanks for the heads up. Just filed at ANI where I imagine this will be quickly handled. Pinged you there out of courtesy in case you want to add anything re: block evasion NicheSports (talk) 21:03, 12 January 2026 (UTC)[reply]

Or not so quickly. Confused by this one NicheSports (talk) 23:57, 13 January 2026 (UTC)[reply]

I assume people are waiting either for them to respond or to start editing again, they haven’t been active since the 9th Kowal2701 (talk) 00:03, 14 January 2026 (UTC)[reply]

On a related issue, Is there a quick way to get the list of articles edited by them needing cleanup? (I mean, without creating it manually) --Friniate ✉ 00:31, 14 January 2026 (UTC)[reply]

For the enwiki pages they've created, see here [82]. I was going to AfD/prod some of these once the ANI thread was resolved.

Also noting their global contribs [83] they've been active more recently at es.wiki, porting over many of the same LLM-generated articles. Does this merit a global lock? NicheSports (talk) 00:36, 14 January 2026 (UTC)[reply]

@NicheSports Thanks, but I meant a list of all edited articles like the subpage created by @Fifteen thousand two hundred twenty four. Is there a quick way to have it or must it be done manually?

I'm going to ask to a steward if it's possible to apply a global lock, but I'm not sure if it will be done as an ANI thread on enwiki is still ongoing... --Friniate ✉ 15:28, 14 January 2026 (UTC)[reply]

Yeah I think requesting a global lock is a good idea, they're active at es.wiki and wikidata (though I personally can't tell if their edits are unconstructive). The tables are created by using User:DVRTed/AINB-helper script Kowal2701 (talk) 15:34, 14 January 2026 (UTC)[reply]

@Kowal2701 Thank you, I'll have a look at it! --Friniate ✉ 16:10, 14 January 2026 (UTC)[reply]

This Spanish language article they created on 11 January [84] is a word-for-word translation of the likely LLM-generated Letter from Leonardo da Vinci to Ludovico Sforza. Given their extensive use of machine-translation on it.wiki, it is probably safe to conclude this was a machine translation. NicheSports (talk) 16:14, 14 January 2026 (UTC)[reply]

(As I thought, the steward I reached confirmed that it's preferable to wait for the conclusion of the ANI procedure here, in order not to overcome local community consensus). --Friniate ✉ 16:23, 14 January 2026 (UTC)[reply]

I didn't use AINB-helper for this one, though that can be a useful tool. The list was fetched using CCI's contribution surveyor (uncheck minor edits, set a relevant date range and a large negative bytes value), then formatted. If an editor wants a list of edited articles without creating a new subpage, then contribution surveyor is a more useful tool. fifteen thousand two hundred twenty four (talk) 17:40, 14 January 2026 (UTC)[reply]

Tracking subpage created and is accessible at the top of this report, 38 articles appear to warrant further review. fifteen thousand two hundred twenty four (talk) 02:01, 14 January 2026 (UTC)[reply]

They have now returned to enwiki and are attempting to G7 several of their article creations, with the edit summary "Italian LTA", which I suppose they are per Friniate, and trying to link to this thread. See [85]. They are also updating the tracking page [86] (sometimes incorrectly). Not sure how to interpret this. NicheSports (talk) 13:54, 14 January 2026 (UTC)[reply]

Have asked them to engage at ANI, but they appear to be trying to aid in the clean-up? Kowal2701 (talk) 15:05, 14 January 2026 (UTC)[reply]

Seems like it? Thanks for reaching out to them. Some communication would be helpful. NicheSports (talk) 15:13, 14 January 2026 (UTC)[reply]

Speaking from experience I tend to have a much more malignant interpretation of his behaviour, he has an history of meddling with cleanups also of other users' edits. I think it's also likely that he's continuing to edit with temporary accounts... --Friniate ✉ 15:24, 14 January 2026 (UTC)[reply]

For example like this one.... --Friniate ✉ 15:30, 14 January 2026 (UTC)[reply]

Or this one... --Friniate ✉ 15:32, 14 January 2026 (UTC)[reply]

As I thought, they for example tagged Giovanni Battista Ramusio as completed, without reverting themeselves... --Friniate ✉ 15:36, 14 January 2026 (UTC)[reply]

@Friniate Anachronist took care of things on enwiki [87]. If you want to do anything on the cross-wiki side I'll leave it to you; re-requesting a global lock makes sense to me but I'm out of my depth here. I looked further into es.wiki and since January 7th they've made ~100 edits there [88], many but not all of which seem to be machine translations of their recent edits to enwiki. Examples: [89][90] :: [91][92] :: [93][94], etc. NicheSports (talk) 07:02, 16 January 2026 (UTC)[reply]

~~It's only a partial block from article creation, so not entirely remedied.~~ (misread the block log) fifteen thousand two hundred twenty four (talk) 19:24, 16 January 2026 (UTC)[reply]

They are blocked from creating or editing articles. PackMecEng (talk) 21:58, 16 January 2026 (UTC)[reply]

@Friniate pinging you here per our discussion at a different talk page. I saw this comment from @Ca yesterday and thought they might be able to help delete this LTA's LLM creations. From xtools [95] Camoz created 36 articles in the LLM-era (a dozen of which are already deleted); from what I have seen, all of these are likely unreviewed LLM-generated output and many can probably be deleted at AfD per WP:NEWLLM as long as they have not been subsequently materially edited by another user. Some of them are pretty egregious, such as List of Carmina Burana. If I weren't traveling I would help AfD some of these but I thought Ca might be willing to take a look. NicheSports (talk) 03:57, 25 January 2026 (UTC)[reply]

TBC elsewhere

Could start a thread at WP:VPI for brainstorming a better process? Though this might need to wait until we get a proper policy at WP:NEWLLM. You might be better off messaging an admin on their talk page, outlining the warnings etc. and asking for a formal warning from them? Kowal2701 (talk) 17:27, 9 January 2026 (UTC)[reply]

I've been thinking about this for a while and have a few ideas. My two basic ideas are 1) establishing that denials of unambiguous LLM use (a high bar, which will mostly apply when there are multiple talk page comments establishing that a user is a non-native English speaker) should be an automatic, typically indefinite, mainspace block, which the user can appeal by acknowledging the LLM use and accepting a complete ban on LLM editing assistance of any kind including LLM-powered "grammar checkers" 2) asking a few admins who are willing to sanction for LLM use to be more active at AINB. Maybe we do the first round at WT:AIC or even on someone's talk page to keep things from going off the rails? I would like to get 1-2 admins' input as well, Chaotic Enby for sure and I was thinking Star Mississippi given Star has handled a lot of the LLM blocks at ANI. Where do you think is best? NicheSports (talk) 17:44, 9 January 2026 (UTC)[reply]

That sounds great, especially the second one. Either ping people to WT:AIC under the premise that you’ve seen them be active on this, or message one of them asking if it’s a good idea (I’m not the best person to ask tbh). If the responses are positive, then a message at WP:AN asking admins to be a little active here probably isn’t a bad idea. Kowal2701 (talk) 18:25, 9 January 2026 (UTC)[reply]

Not really comfortable with a block criterion applying mostly to non-native English speakers, at the very least it is bad optics and may easily seem discriminatory. Gnomingstuff (talk) 20:03, 10 January 2026 (UTC)[reply]

The point isn't about non-native English speakers but identifying denials of unambiguous LLM use. There are other criteria that would apply. I'm also not looking for this to be a PAG as much as general guidance, with the goal of reducing time at ANI. Anyways, I'll open up a separate conversation, likely at WT:AIC, in the next few days. NicheSports (talk) 23:57, 10 January 2026 (UTC)[reply]

(NicheSports—Kowal2701—Fifteen thousand two hundred twenty four—PackMecEng) I started to ask for G15 on two articles created by Camoz (List of Carmina Burana and Bernardino Vázquez de Tapia, but since they are a lot, I prefer to write here before finishing the job: do we all agree to ask for G15 for all the articles created by them? --Friniate ✉ 18:54, 22 January 2026 (UTC)[reply]

@NicheSports sorry, I made a mistake with the ping so it probably didn't arrive to you. --Friniate ✉ 18:56, 22 January 2026 (UTC)[reply]

One of the three specific G15 criteria must be met before G15 can be requested, this means there needs to present either LLM communication, non-plausible hallucinated references, or nonsensical citations. Even if an editor is blocked following a discussion, unless there was consensus in that discussion, blanket deletion via G15 is not a default option.

Other remedies to consider in addition to G15 would be to draftify the article and leave a descriptive talk page message as to why, AfD, tagging the article, or fixing the article. fifteen thousand two hundred twenty four (talk) 19:14, 22 January 2026 (UTC)[reply]

@Fifteen thousand two hundred twenty four So do I have to write in every G15 request why I think that the article was created by an LLM? I'm certainly not going to fix them, as for the tagging I think that it'd just postpone the problem... Maybe G15 for clear-cut cases and draftification where we have reasonable doubts?

PS: I didn't suggest a blanket deletion for all the articles created by them, but only for those already listed in the clean-up page. --Friniate ✉ 20:44, 22 January 2026 (UTC)[reply]

Yes, the page must exhibit one of the three strict G15 criteria before being tagged, and which of those three criteria are best indicated when tagging. Performing a descriptive draftification, either by leaving an appropriate tag or talk page message along with a good edit summary, is often a valid option when G15 is not met.

To clarify, the tracking list contains articles which should be reviewed further, some of the listed pages may need no action taken at all, in which case the unnecessary status should be set. fifteen thousand two hundred twenty four (talk) 20:56, 22 January 2026 (UTC)[reply]

Agreed with fifteen's explanation. I try not to G15 nominate anything that is marginal. I was planning on AfD-ing List of Carmina Burana, as I'm pretty confident that one would be deleted at AfD. I am traveling for work but can look through their other creations for PROD/AfD candidates when I'm back. Anything that isn't deleted can safely be draftified per the machine-generated clause of WP:DRAFTREASON, which is much looser than the G15 criteria, as long as the article also meets the basic rules outlined in WP:DRAFTNO. NicheSports (talk) 22:09, 22 January 2026 (UTC)[reply]

@NicheSports Ok, you have much more experience than me, if you are going to open AfDs I'll leave it to you then, I'll only check the articles which don't need AfDs or to be draftified.. --Friniate ✉ 22:53, 22 January 2026 (UTC)[reply]

By the way, I think that TA as this were most likely always them, so we should pay attention to possible socks. --Friniate ✉ 23:24, 22 January 2026 (UTC)[reply]

Agreed that looks like them. Probably worth adding a few pages to my watchlist... fifteen often looks out for socks as well. Thanks again for letting us know about the cross wiki behavior btw. NicheSports (talk) 23:29, 22 January 2026 (UTC)[reply]

No issue, thank to you for all the work you've done, a wikiproject like this is really cool XD --Friniate ✉ 23:35, 22 January 2026 (UTC)[reply]

They've done a lot of logged out editing as TAs, more than I care to list. For those with TAIV the majority seem to be under the /20 of ~2026-92236, and most recently there's ~2026-29058-7 which is on a different network. Will monitor for a month or so. fifteen thousand two hundred twenty four (talk) 00:22, 23 January 2026 (UTC)[reply]

Edits from User:OrlovskyAtlas

This account was created in November and since then has completely re-written (or attempted to re-write) dozens of articles, mostly in the financial space and on contentious topic WP:CT/IRP. They mostly do not include edit summaries, but when they do they seem obviously auto-generated. Despite receiving numerous warnings for their edits, they have yet to reply to any of them, which makes me further suspicious. There do appear to be minor human-generated edits in-between these massive rewrites, but WP:AISIGNS abound in the larger edits. Most suspicious of all is that several of these extensive re-writes occur just minutes apart (example: 1 and 2); a completely impossible timeline for a human. RunningOnBrains^(talk) 17:53, 9 January 2026 (UTC)[reply]

The user is still engaging in implausible rewrites. Reported at ANIBoynamedsue (talk) 07:07, 12 January 2026 (UTC)[reply]

Courtesy link: WP:ANI § OrlovskyAtlas: Constant use of AI despite warning. fifteen thousand two hundred twenty four (talk) 16:00, 12 January 2026 (UTC)[reply]

I've reverted a handful. Sarsenet•he/they•(talk) 10:45, 12 January 2026 (UTC)[reply]

Suspected LLM use in articles created by MisawaSakura

Came across a few new articles by an extended confirmed user MisawaSakura (talk · contribs · deleted contribs · logs · filter log · block user · block log) that showed some signs of LLM use. I draftified them and asked the creator on their user talk page if they'd used LLMs. They repeatedly dodged the question and eventually ragequit. More background here. Based on Lee Vilenski's suggestion I'm bringing this up here since who knows how many more LLM-generated pages there could be that I didn't scroll down far enough to catch. —pythoncoder (talk | contribs) 18:53, 9 January 2026 (UTC)[reply]

That's quite a lot of mainspace page creations.[96] ~Anachronist (who / me) (talk) 05:31, 16 January 2026 (UTC)[reply]

@Pythoncoder: Kristina Zlateva, that you moved to draftspace on account of its likelihood of being LLM-generated, had several sentences removed but otherwise underwent no fundamental textual rewrite. User:MisawaSakura then submitted it to AfC where it was accepted the following day by User:MurielMary. So unfortunately it still contains the original LLM writing (which a couple of checks have confirmed). —Fortuna, imperatrix 13:05, 17 January 2026 (UTC)[reply]

RfC of possible interest to this project

Image use policy - prioritizing images used by reliable sources when using AI-generated imagery? — Rhododendrites ^talk \\ 20:21, 9 January 2026 (UTC)[reply]

Jībanmṛta and Esyms

Jībanmṛta (talk · contribs) made a large number of LLM-generated edits before they were globally locked as a sockpuppet. Most of these edits fall within the South Asia contentious topic. They had a tendency to add large volumes of densely referenced text in quick succession at a pace that would not be possible if researched and written manually, which was frequently followed by the addition of wikilinks in a separate edit afterward. Examples include:

Terrorism in Bangladesh: 4 November 2025 (18:06, 18:41, 18:57, 19:24, 19:35); 5 November 2025 (05:43, 06:14, 06:51, 06:59)
Sindoor: 15 November 2025 (17:49, 17:58, 18:14, 18:23, 18:25, 18:38)
Ram Mandir Dhwajarohan: 28 November 2025 (09:54, 10:05, 10:17, 10:36, 10:43, 11:25)

Jībanmṛta also participated in the GOCE December 2025 Copy Editing Blitz, in which they claimed to have copyedited articles that are thousands of words long in a matter of minutes. For example, in Special:Diff/1327817371, they removed 4,054 characters from the Sichuan opera article 13 minutes after placing the {{GOCEinuse}} tag to indicate that they had started the copyediting process.

As a sockpuppet of a previously blocked and banned account (CosmLearner), Jībanmṛta's edits can be reversed without needing to provide any further justification per the block evasion (WP:BE) and ban evasion (WP:BRV) policies, as long as at least one of the policies is cited in the edit summary. — Newslinger talk 13:11, 14 January 2026 (UTC)[reply]

Esyms (talk · contribs) has also been globally locked as a sockpuppet of CosmLearner (talk · contribs), the same account that Jībanmṛta is a sockpuppet of. The edit history of User:Esyms/sandbox shows Esyms starting with unsourced LLM-generated content and adding citations to it afterward. — Newslinger talk 15:08, 14 January 2026 (UTC)[reply]

I see that Esyms had been participating in the GOCE January 2026 backlog elimination drive prior to being identified as a sockpuppet, and supposedly made this copyedit to the TV Brasília article in seven minutes. Dhtwiki, as the lead project coordinator of GOCE, is there any action you would like to take in response to this issue? — Newslinger talk 04:35, 15 January 2026 (UTC)[reply]

After looking at a couple of Esyms's copy edited articles, it looks like they were doing good work for the most part. Their "TV Brasilia" edits, which you have reverted, were not all to my liking, but it seemed to be helpful work done in good faith. If you are required, without exception, to undo all of their work, I will probably have to spend some time rescuing much of it. Dhtwiki (talk) 09:48, 15 January 2026 (UTC)[reply]

Thank you for taking a look. Esyms is the 11th sockpuppet listed in Category:Wikipedia sockpuppets of CosmLearner, so I will have to disagree with you on the presence of good faith. Personally, I don't think a sockpuppet who asked an AI chatbot to copyedit articles, and then pasted the contents of the output into Wikipedia, should be listed on the leaderboard, but that is a decision for the GOCE WikiProject and I am simply relaying the information to you. This particular incident is a dual cleanup case because it involves both LLMs and sockpuppetry: not all of Jībanmṛta's and Esyms's edits need to be reverted, but any edits that are not obvious improvements are likely to be reversed. — Newslinger talk 11:55, 15 January 2026 (UTC)[reply]

How can you tell that Esyms really did use AI to do their copy edits? The edits were both observant of MOS (placing dashes in date ranges) and then surprisingly unobservant (rewriting within a quote), but that could be true of purely human copy editing. Of course, Esyms is not going to receive barnstars, nor long remain on the January drive leaderboard. Dhtwiki (talk) 22:51, 15 January 2026 (UTC)[reply]

In Special:Diff/1332237323, Esyms created a user script with the comment "Grok AI been used in most of the places". In their next edit five minutes later, they made significant revisions to the script with the edit summary "AI rewritten". These two edits show that Esyms has access to an LLM (Grok) that would be capable of making the copyedits they submitted, and that Esyms has used this LLM to make edits on Wikipedia. The Jībanmṛta and Esyms accounts were operated by the same individual (CosmLearner), and as I mentioned in my initial comment and my second comment, the timing and order of these accounts' edits to article content are indicative of LLM use.

CosmLearner's other sockpuppets also have a long history of LLM abuse in Wikipedia articles. For example, the most recent edits in Special:Contributions/B'Desh-In Outlook show the addition of tens of thousands of characters of article content within minutes, which is even more ridiculous considering that it was done with the mobile web interface. When we look at the comments that were actually written by the same individual without LLM use, such as Special:Diff/1260009187, Special:Diff/1260433296 (edit summary), and Special:Diff/1256582596, we see a level of English proficiency that would be insufficient for copyediting articles on the English Wikipedia. I don't believe that Esyms/CosmLearner has the ability to perform the GOCE-related copyedits by hand.

Thanks for confirming that Esyms won't receive recognition for the January drive. That was my main concern when I pinged you, because it's not my place to make this type of decision. I recognize that the TV Brasília article does need to be copyedited again, and that many of the changes in Special:Diff/1331321872 were valid, so I will take some time to copyedit the article myself. — Newslinger talk 12:30, 16 January 2026 (UTC)[reply]

Thank you for the lengthy explanation and for seeing that a blanket rollback of Esyms's edits is unnecessary. Dhtwiki (talk) 05:54, 17 January 2026 (UTC)[reply]

I have checked the editor's 11 other articles, other than TV Brasilia, that Esyms copy edited for our drive and found very little that was wrong. If their work was done with an AI tool, it seems to be a very good tool for that kind of work. Dhtwiki (talk) 07:35, 24 January 2026 (UTC)[reply]

Hmlarson

Hmlarson (talk · contribs) is an experienced and autopatrolled editor who tripped 1325 (hist · log) with their creation of PWHL Takeover Tour. If they are using LLMs, the use is occasional, they may subsequently edit some of the content, and I do not think it predates November 2025. See this edit [97] from 2023 which was definitely not LLM-generated; compare to the style of the recently-created Kimbra Walter. What seems clearer is that Hmlarson is creating articles with material source-to-text integrity issues; pretty much the entire Purpose and objectives section fails verification, and is the type of corporate bulleted promo you'd expect from an LLM. Unfortunately, the conversations at Talk:PWHL Takeover Tour about these issues did not go well, with aspersions I was operating a multiple accounts and no engagement with the potential issues raised [98][99]. I placed an ai-generated tag on the article, which they removed [100]. I ran the article through Earwig [101] and found some CLOP (nothing remotely worthy of CCI), but 945 girls' hockey players participated in clinics and meet-and-greets with PWHL athletes is copied word for word from [102].

Snippets from recently-created articles

PWHL Takeover Tour

The Purpose and objectives section almost entirely fails verification as noted above
Talk:PWHL Takeover Tour § Potential WP:V issues documents additional issues
The initiative serves multiple purposes: growing the sport's fanbase, testing potential expansion markets... the bolded bit is SYNTH at best; the article neither mentions nor implies testing markets. This is sourced to a primary source, with support for these claims coming from a quote from a PWHL exec [103]
The rising interest in the PWHL extends beyond its current cities, evidenced by a new single-day ticket sales record... this does verify, but to a primary and promotional source [104]

Kimbra Walter

Large and undue Philanthropy section with LLM-style formatting. Some promo: The organization operates on core values of equity, opportunity, and sustainability.
Walter is a practicing attorney who has dedicated much of her career to philanthropic endeavors. fails verification and is sourced to a primary source [105]
The trophy, designed by Tiffany & Co., is made of sterling silver, weighs approximately 35 pounds, and stands 24 inches tall. completely fails verification, sourced to [106]

NicheSports (talk) 04:44, 16 January 2026 (UTC)[reply]

I did notify Hmlarson about this thread. I'm hoping we can discuss here; I respect their commitment to the articles they work on and they have clearly done a lot of great work on Wikipedia. I would have preferred to do this at a talk page but after the interaction at Talk:PWHL_Takeover_Tour I figured it was best to bring here. NicheSports (talk) 04:49, 16 January 2026 (UTC)[reply]

From a quick glance PWHL Takeover Tour#Format also fails verification.

Given that their only response has been to call you a sock the odds of interaction here seem slim, but hope springs eternal. Zygmeyer (talk) 05:12, 16 January 2026 (UTC)[reply]

the conversations at Talk:PWHL Takeover Tour about these issues did not go well – No kidding, the immediate and repeat followup condescension and bad faith musings are egregious.

I've looked at another one of their recent creations, 2021–22 PWHPA season, and it's so full of LLMisms and unsupported OR it's difficult to convey in a concise manner.

The Season statistics and results section is baffling from a human editor perspective, it's an unsourced section saying there is no source for the information, but makes sense if returned from an LLM.
The unsourced Notable players section contains a list with no notable players, just model bulletpoints of the types of notable players like International players from various countries and Younger players developing their skills.
The existance of an Impact and significance section itself is characteristic of model output, and the mostly-unsourced and editorial contents do not dissuade this notion: The season demonstrated the PWHPA's resilience and adaptability ... The successful navigation of the Olympic season and the continued growth of partnerships proved that the association's ...

This goes on and on. I've not bothered checking for source-text integrity, so much of the article is unsourced OR that it's already well past the threshold of the standards an autopatrolled editor should be holding themselves to. Given this and their recent unconstructive responses to good-faith concerns about their creations, I don't think they should have the perm, but that's something a different venue would have to address. fifteen thousand two hundred twenty four (talk) 05:30, 16 January 2026 (UTC)[reply]

Thanks for looking into this. Agreed on the perm, although I have no appetite for a different venue right now. As for the cleanup, I'm optimistic that it won't be that bad (I was worried when I saw 30k+ contribs). I am pretty confident the apparent LLM use started recently. Now that I have the AINB tracker installed I'll create the tracking page tomorrow. NicheSports (talk) 07:43, 16 January 2026 (UTC)[reply]

S-Aura

S-Aura (talk · contribs · deleted contribs · logs · filter log · block user · block log)

I've tagged several galaxy articles created in December 2024 as AI-generated: NGC 1162, NGC 1163, NGC 1164, NGC 1166. They were full of incorrect information including hallucinated references, I've done some work to clean them up but they should probably just be nuked and rewritten. Other pages created by the same user (S-Aura, now blocked) should be looked into. SevenSpheres (talk) 19:01, 16 January 2026 (UTC)[reply]

SevenSpheres, you can request speedy deletion using WP:G15 if you don't fancy rewriting/stubifying them Kowal2701 (talk) 13:57, 17 January 2026 (UTC)[reply]

Sugar Laden

Sugar Laden (talk · contribs · deleted contribs · logs · filter log · block user · block log)

User is creating multiple articles incredibly quickly with non-existent categories and markdown. Need these to either be cleaned-up or deleted. – LuniZunie_(talk) 15:01, 18 January 2026 (UTC)[reply]

sorry maam Sugar Laden (talk) 15:02, 18 January 2026 (UTC)[reply]

Seems like User:~2026-37658-6 is also them per WP:DUCK, as the redirects made by Sugar Laden were converted into AI generated articles by the temp user. – LuniZunie_(talk) 15:13, 18 January 2026 (UTC)[reply]

Brought to ANI but still need cleanup. – LuniZunie_(talk) 15:19, 18 January 2026 (UTC)[reply]

Editor blocked, three pages remain: Keiko Necesario, Gaano Ko Ikaw Kamahal (deleted) and ~~Draft:Pasko Na, Sinta Ko~~ (G15 tag removed for now) are all G15 tagged. fifteen thousand two hundred twenty four (talk) 17:39, 18 January 2026 (UTC)[reply]

Gaano Ko Ikaw Kamahal has been G15'd. SuperPianoMan9167 (talk) 17:57, 18 January 2026 (UTC)[reply]

Rumba Rapids

Hi, I was reading the page for Rumba Rapids and noticed that a large part had been rewritten by AI and pasted into the article complete with a note from the AI. I've read the guide but I'm still not sure whether to just remove the note or revert to the pre AI version (given that it was not very well written before). Many Thanks ~2026-38947-2 (talk) 19:50, 18 January 2026 (UTC)[reply]

@~2026-38947-2

Done, I edit conflicted with @Fifteen thousand two hundred twenty four who removed it, thanks for finding that. HurricaneZeta^C 19:59, 18 January 2026 (UTC)[reply]

And to answer your question ~2026-38947-2, if there is clear evidence that an edit is LLM-generated and unreviewed, a revert is appropriate. fifteen thousand two hundred twenty four (talk) 20:00, 18 January 2026 (UTC)[reply]

Also the entire ride section seems to be original research and was written unencyclopedically so I just removed it. HurricaneZeta^C 20:04, 18 January 2026 (UTC)[reply]

BassiStone cleanup

BassiStone (talk · contribs) was community banned for using AI/LLM on various articles over the past few months. To make a long story very short, I only got involved in this because, as a regular on the Britney Spears article, I checked their contributions to Spears's 1999-2001 personal life section and on Swan Lake and some of these edits were in fact created via AI-generated information.

Given that, along with the fact that any edit by a banned user (good or bad) would be reverted (I have reverted some of them myself), I have been considering a potential cleanup project to weed out any AI/LLM issues on the articles the user has edited as a whole. sjones23 (talk - contributions) 17:25, 19 January 2026 (UTC)[reply]

along with the fact that any edit by a banned user (good or bad) would be reverted – The shortcut name of WP:BANREVERT is a little misleading. Banreverts only apply to edits made in violation of a ban or block, BassiStone's edits do not qualify since they were made prior to any ban or block. fifteen thousand two hundred twenty four (talk) 20:10, 19 January 2026 (UTC)[reply]

If there was a ban evader in general, we would revert them on sight and report it.

However, as what you pointed out, BassiStone's edits don't count since he’s not currently a ban evader. With the ban evader statement out of the way, one of my main points in this cleanup suggestion is that we should consider resolving the AI/LLM issues. At this point, should we file a cleanup request? sjones23 (talk - contributions) 22:21, 19 January 2026 (UTC)[reply]

Tracking subpage created, 48 articles await closer review. Only edits after 23 June 2025 were considered as this appears to be when more obvious LLM use began [107]. fifteen thousand two hundred twenty four (talk) 01:59, 20 January 2026 (UTC)[reply]
Thanks. As a side note, the Britney Spears article AI issue was already dealt with a few weeks ago. sjones23 (talk - contributions) 06:05, 20 January 2026 (UTC)[reply]

Goldenpen1 at ANI

Goldenpen1 (talk · contribs · deleted contribs · logs · filter log · block user · block log)

Currently being discussed at ANI: Wikipedia:Administrators' noticeboard/Incidents#Goldenpen1 - LLM-generated content and fraudulent sourcing. A couple hundred edits but mostly to the same few articles. Thebiguglyalien (talk) 🛸 03:58, 26 January 2026 (UTC)[reply]

Tracking subpage created. Cleanup complete. fifteen thousand two hundred twenty four (talk) 21:17, 26 January 2026 (UTC)[reply]

Order of the Knights of the Fiery Cross

Just removed an ChatGPT-generated image from this article; a lot of the text seems likely to be AI-generated as well. Omphalographer (talk) 03:59, 29 January 2026 (UTC)[reply]

v t e Noticeboards
Wikipedia's centralized discussion, request, and help venues. For a listing of ongoing discussions and current requests, see the dashboard. For a related set of forums which do not function as noticeboards see formal review processes.
General	Administrators Main Incidents Bots Bureaucrats Centralized discussion Closure requests Education Interface admins Main Page errors Open proxies Volunteer response team Oversight User permissions
Articles, content	AI Cleanup Bad image list Biographies of living persons Copyrights Questions on media Problems Dispute resolution External links Fringe theories Neutral point of view Original research Pending changes Reliable sources Resource requests Scalable vector graphics Spam Blacklist Whitelist Style Titleblacklist Translation
Page handling	History merges Mergers Splits Moves Protection Importation XfD Articles Redirects Categories Templates Files Miscellany Undeletion Deletion review
User conduct	Conflict of interest Contributor copyright Edit warring and 3RR Sanctions Personal restrictions General sanctions Contentious topics Sockpuppets Usernames (Requests for comment) Vandalism
Other	Arbitration Committee noticeboard Requests Enforcement Edit filters Requested False positives Questions Help desk Teahouse Reference desk New articles Requests for comment Village pump Policy Technical Proposals Idea lab WMF Miscellaneous WikiProject proposals Discussions for discussion
Category:Wikipedia noticeboards