Jump to content

Talk:Generative engine optimization

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Vendors of Generative engine optimisation tools

[edit source]

The two vendors of tools listed on this page seems rather limited. I think it would benefit from a list of current vendors, but before I do that, I wanted to get some feedback. It would be good to look at categorising as different tools have different features. I don't want to make this a catalog of vendors, but I do think the current article is inadequate. Any thoughts? MikeMaynardUK (talk) 14:18, 2 January 2026 (UTC)[reply]

I agree with you that the article would benefit from more discussion of companies that are successfully implementing GEO. Semrush is developing GEO, its primary strength and focus are still SEO. I dont think it is the best example without mentioning that. I recently added Evertune AI, which is one of the more successful and innovative companies in this field, but that content was removed as "spammy". At the moment, the article itself is still very underdeveloped—it’s rated C-class by who?, has just a few sources, and contains several “citation needed” tags.
I don’t have the capacity right now to create a comparative chart, but I can try to rewrite that section in a more neutral, well-sourced way. You’re welcome to do the same. Which companies are you most familiar with? I would also consider Profound as a candidate for inclusion, I would think there is sourcing for that. Kristopher9 (talk) 19:11, 16 January 2026 (UTC)[reply]
I think a chart would be dangerous. Probably best to just list a few examples. I don't think Wikipedia is best served as a place where people can comparison shop. Unless I'm misunderstanding the premise. Dflovett (talk) 04:06, 30 January 2026 (UTC)[reply]
Yes I agree now after watching the article, It is a spam magnet and the chart would be misused with every little startup entering their name. Kristopher9 (talk) 14:37, 30 January 2026 (UTC)[reply]
I agree. I think it should mention Semrush, Ahrefs, Profound, Scrunch, and Peec. But honestly, I see an argument for it naming even more than that. Mostly drawing on my industry knowledge paired with where I see it being mentioned. Dflovett (talk) 19:48, 5 February 2026 (UTC)[reply]
I was going to put in Evertune AI, but the spam was just so bad that I thought to not start something. Bluefish is an ok example, though it looks like it their article was deleted, and I also was thinking Profound. Semrush needs to be explained as it is still primarily SEO, but when I tried to qualify which product they had that was GEO, it was taken as spam. I went in and tried to source everything without naming anyone, just to help, but the editors have a point in deleting everything. It is not the time. Kristopher9 (talk) 01:02, 6 February 2026 (UTC)[reply]
That's fair. This is only further complicated by the AIO and AEO articles that also exist. Dflovett (talk) 04:20, 6 February 2026 (UTC)[reply]
Hi, I professionally research Generative Engine Optimization within a EU uni setup and was working as SEO engineer for a decade in the EU. None of these companies qualify for GEO: 1) Not GEO specific (except maybe Profound and Peec, Inlinks could deserve a mention too) 2) GEO is still developing as a discipline 3) Hardly a marketing technique anymore. Emilija gjorgjevska (talk) 15:56, 11 April 2026 (UTC)[reply]

There are a few out there, the current examples are misleading, as Semrush is mainly SEO focused. A more up to date or accurate representation would be very helpful.--~2026-36833-9 (talk) 16:36, 20 January 2026 (UTC)[reply]

I tried to make a responsible, sourced edit about this, but it was reversed. The problem is that new startups are constantly inserting their names into the article, so the good content is getting reverted along with the bad. My rewrite explained the role of Semrush, which was developing SEO, and then added a couple of the more prominent dedicated GEO companies, along with the already-listed Bluefish AI. However, the editor seemed to think I was promoting Semrush and reverted it. This is an important topic that is currently very messy. Kristopher9 (talk) 17:04, 20 January 2026 (UTC)[reply]
I think this is where third-party coverage is important. Like, a few good reliable articles that have no association with any of the tools that also list all the tools are going to be the best way to determine what should be listed. I don't think it makes sense currently for Bluefish AI and Semrush to be the only two listed, but I'm not ready to make the edits yet. Dflovett (talk) 16:54, 5 February 2026 (UTC)[reply]

Search Engine Land as a source?

[edit source]

I have some questions about using SEL as a source post-Semrush acquisition. Especially in an article where Semrush was mentioned previously and could be mentioned again. However, maybe it depends on who the author of the given article is. Curious what other editors think. Dflovett (talk) 14:43, 11 February 2026 (UTC)[reply]

@Dflovett: I don't think that's any reliable, but it's way better than the spam sources that are added every now and then. Children Will Listen (🐄 talk, 🫘 contribs) 13:37, 26 February 2026 (UTC)[reply]
I completely agree. These pages are spam magnets. Dflovett (talk) 14:39, 26 February 2026 (UTC)[reply]
SEL is super inapporpriate here and not a valid source for GEO. Arxiv papers who are not peer-reviewed yet still employ scientific method are a much better resource on this. Please fix. Emilija gjorgjevska (talk) 15:58, 11 April 2026 (UTC)[reply]
Planning to add the remaining Practitioner tactics subsections soon. Given the concerns raised here about SEL as a source, I'm going to avoid SEL refs entirely and stick to arXiv and academic sources instead. Rickvlack (talk) 13:36, 27 April 2026 (UTC)[reply]
arXiv is mostly self published stuff and should generally be avoided - unless you are sure the paper has also been published elsewhere. MrOllie (talk) 13:38, 27 April 2026 (UTC)[reply]
Good to know. What would you suggest using instead? Most of the academic work is still at preprint stage. Rickvlack (talk) 15:10, 27 April 2026 (UTC)[reply]
Wikipedia shouldn't be leading the charge, and there aren't any deadlines. If most of the work is still at preprint the sourcing guidelines would suggest that we should wait a bit and let the publishing process run. MrOllie (talk) 15:16, 27 April 2026 (UTC)[reply]
I fundamentally disagree, since this can take years until we product more solid work. At the moment, the number of peer reviewed work on the topic of GEO is < 10. I think the proposed approach is wrong. Also, there are many fields, not just in SEO/GEO, that have better resources on arxiv for specific scientific niches, instead of top notch conferences.
If I am not wrong, one of the most revolutionary papers like Attention Is All You Need was not a peer reviewed article. Therefore, this argument is not an argument at all.
Cheers,
E ~2026-25759-22 (talk) 15:21, 27 April 2026 (UTC)[reply]
this was me btw, super weird that I got flagged as a temporary account -.- Emilija gjorgjevska (talk) 15:22, 27 April 2026 (UTC)[reply]
Following policies such as WP:V, WP:NOR and WP:RS often means that Wikipedia is behind the bleeding edge, sometimes by years. That isn't a problem, that is specifically how Wikipedia is designed to work. We summarize, old, boring settled science, we're not a news outlet reporting on the latest revolutionary paper. In the case of 'Attention is All You Need', like most truly revolutionary papers, it generated a lot of citations and secondary coverage. That's the kind of stuff Wikipedia uses. MrOllie (talk) 15:31, 27 April 2026 (UTC)[reply]
precisely that, and when we are already discussing this point, please use it to remove SEL as as reference, since it is portraying exactly that.
Also, i disagree that waiting years for science to flourish here is the right approach. This is a field in its infancy and deserves better than us being followers in the fields, rather than leaders. Ofc, while being in line with how science works, because arxiv preprints are totally legit :) Emilija gjorgjevska (talk) 15:33, 27 April 2026 (UTC)[reply]
You're welcome to disagree with how Wikipedia is architected, and you can try to get the policies changed at WP:VPP. But until you do, we should follow the existing policies. MrOllie (talk) 15:37, 27 April 2026 (UTC)[reply]
Not trying to change policy, please don't misquote me. Still, helpful to know that something like this is possible, thanks. Emilija gjorgjevska (talk) 15:37, 27 April 2026 (UTC)[reply]
I find it astonishing _how is it possible_ that we use one approach for SEL, yet, abandoning this same exact logic when it comes to arxiv. Simply crazy, I can't understand this sorry. Emilija gjorgjevska (talk) 15:40, 27 April 2026 (UTC)[reply]
SEL isn't a great source either, never said it was. MrOllie (talk) 15:45, 27 April 2026 (UTC)[reply]

Balkanization of Discovery

[edit source]

One of the article's citations has the provocative "Is SEO dead" title. Would it be more balanced to describe the current SEO, AEO, GEO milieu as entering a balkanization phase where multiple discovery/visibility strategies yield different coverage and results?Data-and-Facts (talk) 13:43, 25 February 2026 (UTC)[reply]

I think "SEO Is dead" is definitely not a narrative that should be included in this, as that's something that people have been saying for decades and will continue saying for decades. But if the citation in question is notable, it's okay to use it. I agree with you on the balkanization phase theory but not sure how best to capture that on here. I have attempted that on the search engine optimization article but that admittedly needs work as well. Dflovett (talk) 20:35, 26 February 2026 (UTC)[reply]
SEO is dead is super wrong and that narrative is not scientifically supported through a reproducible, scientifically valid method. This is not a forum, this is a Wikipedia page and references like these should be immediately removed. Emilija gjorgjevska (talk) 16:02, 11 April 2026 (UTC)[reply]
[edit source]

I've redirected answer engine optimization and artificial intelligence optimization to this article since they appear to be the same concept referred to using different names. AI SEO remains since it seems to be talking about enhancing existing SEO tactics with AI tools, not to optimize content for LLM consumption and retrieval. Children Will Listen (🐄 talk, 🫘 contribs) 13:50, 26 February 2026 (UTC)[reply]

How did you choose this one as the survivor? I did notice the pageviews were higher on this than the other two and the search volume does look higher for "generative engine optimization" than the other two. I do think the answer engine optimization article itself was strong in some places and I will likely make a few edits to this article to strengthen it (artificial intelligence optimization was a mess and unsalvageable IMO.) Dflovett (talk) 14:42, 26 February 2026 (UTC)[reply]
This article has the greatest citation to article content density, and is arguably more active. Children Will Listen (🐄 talk, 🫘 contribs) 14:53, 26 February 2026 (UTC)[reply]
Makes sense. I am going to likely make some edits. Dflovett (talk) 15:25, 26 February 2026 (UTC)[reply]
One more thought. If AI SEO doesn't get redirected into this one, I think it could easily be absorbed into Search engine optimization Dflovett (talk) 21:16, 26 February 2026 (UTC)[reply]

Proposal: add Mechanisms, Practitioner tactics, Terminology distinctions, and Measurement section

[edit source]

Rickvlack (talk) 15:03, 25 March 2026 (UTC)[reply]

I have a few concerns with this. I think you should ensure you have reliable sources before making any edits. This article has been getting spammed a lot. All four of those seem like reasonable sections to me, so it will really be a matter of citing reliable sources and slowly building this out. If you look through the revision history, you can see a lot of people approaching this incorrectly that make for good lessons. Dflovett (talk) 02:25, 27 March 2026 (UTC)[reply]
Thank you for the feedback and for taking a look. Noted on the sourcing. I'll make sure every claim is tied to a reliable third-party source before adding anything, and I'll mark anything I can't source as [citation needed] rather than leaving it unsupported. I'll start with the Terminology distinctions section as it's the most straightforward and has the least risk of sourcing gaps. Happy to share the draft text here first if you'd like to review before it goes live. Rickvlack (talk) 14:03, 30 March 2026 (UTC)[reply]
I wouldn't mind taking a look. My recommendation is always start small. Dflovett (talk) 16:22, 30 March 2026 (UTC)[reply]
Hi Dflovett — the week is up and I'm ready to make the first edit. Taking your advice to start small, I'm proposing to add just the Terminology distinctions section first. Here is the full draft text for your review before I add it to the article:
== Terminology ==
Several overlapping terms describe related practices, and usage varies across practitioners, vendors, and publications. No consensus definition distinguishing these terms had been established in the academic literature as of early 2026, and the terms are frequently used interchangeably in trade and practitioner contexts.[1]
''Answer engine optimization'' (AEO) is sometimes used specifically in reference to systems designed to return direct answers rather than lists of links — such as voice assistants and featured snippet formats — predating the widespread deployment of large language model-based search.[citation needed]
''Large language model optimization'' (LLMO) is used in some practitioner contexts with a narrower focus on influencing a model's parametric knowledge rather than on retrieval-based systems.[citation needed]
''Artificial intelligence optimization'' (AIO) appears in some vendor and agency contexts as a broader umbrella term covering any practice aimed at improving an entity's representation across AI systems.[citation needed]
''AI SEO'' is used when the practice is positioned as a direct continuation of traditional search engine optimization workflows adapted for AI-mediated discovery environments.<[1]
Happy to adjust anything before this goes live. Rickvlack (talk) 16:23, 1 April 2026 (UTC)[reply]
Rickvlack (talk) 16:23, 1 April 2026 (UTC)[reply]
I like this! I could see some small edits but I think you're off to a good start. Terminology is a big part of what's debated and disussed around this.
I would personally drop the em dashes. This doesn't read like you used an LLM to write it but em dashes sometimes lead to accusations of LLM content. Dflovett (talk) 14:06, 2 April 2026 (UTC)[reply]
Good catch and useful to know! I've updated the draft to remove the em dashes. Here is the revised version, happy for you to let me know if you're satisfied for this to go live:
== Terminology ==
Several overlapping terms describe related practices, and usage varies across practitioners, vendors, and publications. No consensus definition distinguishing these terms had been established in the academic literature as of early 2026, and the terms are frequently used interchangeably in trade and practitioner contexts.[1]
''Answer engine optimization'' (AEO) is sometimes used specifically in reference to systems designed to return direct answers rather than lists of links, such as voice assistants and featured snippet formats, predating the widespread deployment of large language model-based search.[citation needed]
''Large language model optimization'' (LLMO) is used in some practitioner contexts with a narrower focus on influencing a model's parametric knowledge rather than on retrieval-based systems.[citation needed]
''Artificial intelligence optimization'' (AIO) appears in some vendor and agency contexts as a broader umbrella term covering any practice aimed at improving an entity's representation across AI systems.[citation needed]
''AI SEO'' is used when the practice is positioned as a direct continuation of traditional search engine optimization workflows adapted for AI-mediated discovery environments.[1]
Rickvlack (talk) 19:19, 2 April 2026 (UTC)[reply]
See if you can find citations. I'll look too. I think this is directionally correct based on my own knowledge of the industry and it's good to have this clarified, but citations will make it so much stronger. Dflovett (talk) 14:20, 3 April 2026 (UTC)[reply]
Thanks for looking. Here is the same draft with citations added for each term. I have kept the [citation needed] tags only where I genuinely could not find a source that maps cleanly enough to the specific claim.
== Terminology ==
Several overlapping terms describe related practices, and usage varies across practitioners, vendors, and publications. No consensus definition distinguishing these terms had been established in the academic literature as of early 2026, and the terms are frequently used interchangeably in trade and practitioner contexts.[1]
''Answer engine optimization'' (AEO) is sometimes used specifically in reference to systems designed to return direct answers rather than lists of links, such as voice assistants and featured snippet formats, predating the widespread deployment of large language model-based search.[2]
''Large language model optimization'' (LLMO) is used in some practitioner contexts with a narrower focus on influencing a model's parametric knowledge rather than on retrieval-based systems.[3]
''Artificial intelligence optimization'' (AIO) appears in some vendor and agency contexts as a broader umbrella term covering any practice aimed at improving an entity's representation across AI systems.[4]
''AI SEO'' is used when the practice is positioned as a direct continuation of traditional search engine optimization workflows adapted for AI-mediated discovery environments.[1]
Rickvlack (talk) 16:21, 3 April 2026 (UTC)[reply]
Keep in mind that I'm not the be-all-and-end-all of this article, but I think this makes enough sense to try to add in. In your edit note say to reference the Talk page. Dflovett (talk) 18:03, 3 April 2026 (UTC)[reply]
I thinl we should add this resource from Addy Osmani: https://addyosmani.com/blog/agentic-engine-optimization/, for context: Addy Osmani is a prominent Irish software engineer, author, and current engineering leader at Google, currently acting as a Director for Google Cloud AI. Emilija gjorgjevska (talk) 17:52, 15 April 2026 (UTC)[reply]
Thanks Dflovett, I'll add it in. Emilija, I had a look at the Osmani piece. It's interesting but it's a personal blog post, even from someone with his profile. I'm not sure it meets the sourcing bar for this article. Happy to be wrong on that, what do others think? Rickvlack (talk) 19:07, 15 April 2026 (UTC)[reply]
Rickvlack (talk) 16:21, 3 April 2026 (UTC)[reply]
Rickvlack (talk) 19:19, 2 April 2026 (UTC)[reply]
Coursera is using GEO as a reference https://www.coursera.org/specializations/generative-engine-optimization-geo. My research program where I do research this uses the same and all arxiv preprints use GEO. Please default to GEO. Emilija gjorgjevska (talk) 16:04, 11 April 2026 (UTC)[reply]
The Coursera and arXiv citations are helpful evidence for this. The Terminology section recently added to the article makes a similar point, that GEO has emerged as the dominant term across practitioner and academic contexts, with AEO, LLMO and others being used as related but subordinate terms. Happy to strengthen that section further if there are additional academic sources worth citing. Rickvlack (talk) 13:24, 13 April 2026 (UTC)[reply]
At the moment, there is no peer-reviewed research that can be shared as a reference and we need to be transparent on that one. However, arxiv is much better compared to blogs, personal opinions or industry pushbacks that do not employ at least some scientific method, share dataset and similar preeliminary approaches.
Some resources from Arxiv can be listed here:
1. https://arxiv.org/search/?query=generative+engine+optimization+geo&searchtype=all&abstracts=show&order=-announced_date_first&size=50
Hope this helps.
Cheers,
E Emilija gjorgjevska (talk) 14:15, 13 April 2026 (UTC)[reply]
The article currently does default to GEO, hence the name, but I think the terminology section is useful and a good addition. Dflovett (talk) 14:18, 13 April 2026 (UTC)[reply]
Aha, I understand better now. You are right then, Balkanization of discovery makes sense as section of context. Even though I would use another word than balkanization (tranparency: I am from the Balkans). Emilija gjorgjevska (talk) 14:21, 13 April 2026 (UTC)[reply]
Oh I don't think "Balkans" or "Balkanization" belongs in here in any literal sense. I think that was another editor's suggestion that was not acted on. Dflovett (talk) 14:27, 13 April 2026 (UTC)[reply]
OK, amazing work, thank you very much for the fast replies :)
Kindly take a look into the other comments like SEO is dead reference, I provided some arguments why I am against.
All the best from my side.
Cheers,
E Emilija gjorgjevska (talk) 14:30, 13 April 2026 (UTC)[reply]
Thanks for the arXiv search link, that's a useful aggregated view of the academic usage. Will take a look at the SEO is dead thread now. Rickvlack (talk) 15:18, 13 April 2026 (UTC)[reply]

References

  1. ^ a b c d e f Cite error: The named reference :2 was invoked but never defined (see the help page).
  2. ^ "Answer Engine Optimization (AEO): The comprehensive guide for 2026". CXL. 27 January 2026.
  3. ^ "LLMO and GEO: What We Know About Optimizing for LLMs and AI". Outerbox. 22 August 2025.
  4. ^ "Large Language Model Optimization (LLMO) explained". Evergreen Media. 12 February 2026.

Sourcing for upcoming sections

[edit source]

@Dflovett: , planning to add the Mechanisms section around April 22 and Practitioner Tactics a couple of weeks after. Noticed Kuru removed some sources on April 19 so I want to check before I post anything. For the practitioner tactics claims about content structure and third-party authority signals, what source types would work? The Aggarwal et al. KDD 2024 paper is already in the article and seems to cover some of it. Happy to hold off until we've agreed on sourcing. Rickvlack (talk) 15:09, 20 April 2026 (UTC)[reply]

@Dflovett just checking in. Still wondering about the SEL sources given Emilija's comment, happy to wait if you need more time to look at it. Rickvlack (talk) 14:04, 22 April 2026 (UTC)[reply]
It didn't cite any reliable sources, so I removed this again. I wouldn't oppose a rewrite once proper sourcing becomes available, though. MrOllie (talk) 15:42, 27 April 2026 (UTC)[reply]
Fair enough. I'll rewrite it with proper peer-reviewed sources before resubmitting. I have the Aggarwal et al. KDD 2024 paper and Li & Sinnamon (ASIS&T 2024, Wiley), both should cover what's needed. Rickvlack (talk) 17:22, 27 April 2026 (UTC)[reply]
One more thing - I'm going to cut the knowledge graph sentence in entity disambiguation. Can't find a solid source for it so easier to just remove it than leave the citation needed tags in. Rickvlack (talk) 15:54, 28 April 2026 (UTC) Rickvlack (talk) 15:54, 28 April 2026 (UTC)[reply]
hey @Rickvlack - apologies for not responding to your recent posts in here. I somehow missed the notifications. I think you've made some good improvements overall. I'm going to give it all a more thorough read. Thank you for not introducing any spam! I think SEL is okay to use in certain contexts, especially because it's one of the few historically reliable sources that's really discussing this in depth. Dflovett (talk) 13:38, 4 May 2026 (UTC)[reply]

Moving items from the lead into new or existing sections

[edit source]

Specifically referring to:

The concept of GEO first appeared in response to generative AI technologies being integrated into mainstream search and information retrieval systems.[4]

Tools are used to monitor how websites and brands are cited, referenced, or incorporated into responses produced by large language models.[5]

Practitioners also measure how often a brand is mentioned in AI-generated answers, which URLs or domains are cited in those answers, and a brand’s share of voice relative to competitors.

@Rickvlack you've been pretty active with improvements. Thoughts on this? Dflovett (talk) 13:50, 4 May 2026 (UTC)[reply]

Thanks for flagging the SEL thing, good to know. On the lead, yeah those sentences probably belong deeper in the article. The origins sentence could work in a background section if that ever gets added. The monitoring and share of voice ones I was planning to fold into Measurement anyway, so I'll work them in when I get there. Rickvlack (talk) 15:03, 4 May 2026 (UTC)[reply]
I might try playing around with an "origins" section, although realistically that might resemble Terminology. Maybe I can find a way to make it work as one section. Dflovett (talk) 18:45, 4 May 2026 (UTC)[reply]

Li and Sinnamon sourcing concerns

[edit source]

I've replaced three citations to Li and Sinnamon (2024) with {{cn}} tags. The full citation being removed:

Li, Alice; Sinnamon, Luanne (2024). "Generative AI Search Engines as Arbiters of Public Knowledge: An Audit of Bias and Authority". Proceedings of the Association for Information Science and Technology. 61 (1). Wiley: 205–217. doi:10.1002/pra2.1021.

The paper is basically a review of how AI tools like ChatGPT, Bing Chat, and Perplexity can show bias, for example, in the tone they use, the businesses they favor, or the regions and sources they rely on. However, it doesn't actually back up the specific claims it was being cited for. Those claims were things like:

  • using the same name consistently across websites helps AI recognize and surface an entity,
  • conflicting descriptions make AI give more cautious or uncertain answers,
  • and structured data markup helps AI tell entities apart correctly.

Those ideas may still be correct and supported elsewhere, but this particular paper by Li and Sinnamon doesn't seem to discuss them.

The proper AIO source for the GEO context is already in the article: Kozinets and Gretzel (2024), "From SEO to AIO," cited later in the Terminology section. but I am leaving this for regular maintainers of the article to consider. AThickerhead (talk) 15:24, 9 May 2026 (UTC)[reply]

@Rickvlack, could you explain why you added the Li and Sinnamon source here and here? I don't see how that source relates to entity consistency or entity disambiguation, or GEO in general. I skimmed through the paper and its main research questions are listed as "RQ1: Are Generative AI responses influenced by the sentiment of the query and topic?" and "RQ2: How do Generative AI search engines establish authority in their responses to queries?" Truthnope (talk) 19:48, 9 May 2026 (UTC)[reply]
Li and Sinnamon look at how AI search engines establish authority across sources, which I thought was relevant to entity consistency. But I take the point that the connection to the specific claims could be tighter. Happy to look for something more direct if that helps. Rickvlack (talk) 01:20, 10 May 2026 (UTC)[reply]
You need to be very careful about how you use sources. One of Wikipedia's core policies is WP:VERIFIABILITY, which means that readers should be able to verify that the statements in a Wikipedia article come from a reliable source. If an article says "Consistent naming, location data, category descriptors, and structured data markup across web properties help generative models identify and distinguish an entity accurately" followed by a citation, readers should be able to find that the cited source actually makes that claim.
Wikipedia articles are written by reading reliable sources and adding information from those sources. It sounds like you're going about this the other way around, that is, you are adding content you want to see to a Wikipedia article and then tacking a citation on at the end, even if the citation does not support the content. You need to make sure that any citations you've used on this article or any others actually support the content that precedes it. If you cannot find a source to support the content, the content should be removed until a source can be found.Truthnope (talk) 02:05, 10 May 2026 (UTC)[reply]
Fair point on verifiability. With Li and Sinnamon I thought RQ2, which covers how AI search engines establish authority in their responses, was close enough to entity consistency to be relevant. Looking at it again I can see the connection isn't direct enough and I should have been more precise. But that's not the same as adding content and hunting for a citation to justify it. I'll be more careful. Rickvlack (talk) 13:24, 11 May 2026 (UTC)[reply]
I don't want to keep harping on this point, but I want to make sure you understand the issue. You shouldn't think that the source is relevant to the content, you should know that the source verifies the content. Reading the source should precede adding the content. If you don't, you risk 1) linking readers to a source that doesn't actually confirm the content preceding it, and 2) misrepresenting a source. Truthnope (talk) 22:28, 11 May 2026 (UTC)[reply]
One other note @Rickvlack - I think @Truthnope has made their point, but generally I like to start with the source and determine what in there is relevant, rather than constructing the narrative and then seeking a source. I think that might make things easier for you. Dflovett (talk) 13:24, 13 May 2026 (UTC)[reply]
Worth saying, the other contributions went through Talk page discussion and sourcing before they went live. The Li and Sinnamon connection was a misjudgment, not how I've been approaching this generally. Rickvlack (talk) 13:27, 11 May 2026 (UTC)[reply]

Mechanisms section removal

[edit source]

Wanted to raise the Mechanisms section that was removed on May 10. The edit summary said Aggarwal doesn't discuss RAG, but the paper is really about how generative engines retrieve and synthesize information from multiple sources, which is what RAG describes. I'd like to restore it but wanted to talk it through here first rather than just reverting. Happy to find a different source if the Aggarwal connection genuinely doesn't hold up. Rickvlack (talk) 14:07, 15 May 2026 (UTC)[reply]

The problem is that if you write a sentence about RAG and add a reference to Aggarwal at the end of the sentence, you are saying that Aggarwal is making this claim about RAG, which is not true. Think of this from the perspective of a reader: if a reader sees this sentence, they may look at the source to learn more about RAG. When they see that Aggarwal does not discuss RAG, then they won't understand where this information comes from in the first place.
If you can find a source that discusses RAG, you can include a sentence about RAG that is validated by that source. If Aggarwal discusses something similar to RAG, you can write a sentence based on what Aggarwal says and use that as a source. But you can't describe something that is only similar to what Aggarwal says and use Aggarwal as a source. You can't use Aggarwal as a source to talk about RAG if Aggarwal does not mention RAG. This is the essence of our previous discussion. Truthnope (talk) 06:23, 16 May 2026 (UTC)[reply]
Fair point on the terminology. Aggarwal doesn't use the term RAG explicitly, but the paper does describe the same process, generative engines retrieving relevant documents from a database and synthesizing responses from those sources. I could rewrite the section using Aggarwal's own language rather than the RAG label. Would that work, or would you rather I find a source that uses the RAG term directly? Rickvlack (talk) 18:41, 16 May 2026 (UTC)[reply]
Your suggestions both work. Another option is that you could say something like "Aggarwal discusses how to optimize around the process by which generative engines retrieve and synthesize data," and then say "This process is referred to as retrieval augmented generation (RAG)" and cite a source that uses that term. This could be considered a form of WP:SYNTH, but an acceptable form of it, where the article is carefully sourced and at most, it says that this process that Aggarwal addresses is named RAG in other sources. Truthnope (talk) 19:32, 16 May 2026 (UTC)[reply]

Request to replace unsourced Factors paragraph with findings from the KDD 2024 GEO study

[edit source]
Not done: your request appears to have been generated by a large language model. Day Creature (talk) 18:54, 24 May 2026 (UTC)[reply]

References

  1. ^ Aggarwal, Pranjal; Murahari, Vishvak; Rajpurohit, Tanmay; Kalyan, Ashwin; Narasimhan, Karthik R.; Deshpande, Ameet (2024). "GEO: Generative Engine Optimization". Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery. pp. 5–16. doi:10.1145/3637528.3671900.
Retrieved from "https://en.wikipedia.org/w/index.php?title=Talk:Generative_engine_optimization&oldid=1357564295"