Google is trying to reclaim AEO and GEO as "still SEO". The data tells a different story.
Google's new guide says AEO and GEO are extensions of SEO. Ahrefs research across 15,000 queries shows 80% of the URLs AI systems cite never appear in Google's top 100. That is not a technical footnote. That is a different game.
5 min read
0:00
0:00
What Google actually published this week
Google's guide says AEO and GEO are not separate disciplines, and that AI-specific tactics like llms.txt are unnecessary. That is what the document says. Who wrote it and why is a second question.
Google released a guide that explicitly labels Answer Engine Optimization and Generative Engine Optimization as "still SEO." The guide advises site owners to ignore llms.txt files, manual content chunking, and special schema markup aimed at AI systems. The message: the fundamentals of search optimization are unchanged, and everything consultants are selling as "AI SEO" is, in Google's framing, unnecessary. Those are the facts as Search Engine Journal reports them. What the guide leaves out is the more interesting story.
Why Google is saying this now
Google is losing queries to AI assistants. A guide that declares "it is all still SEO" is a reframing to keep the market on terrain where Google is strongest. It is not a technical truth.
A guide from Google about AI search is not an independent technical document. It is market communication from a company that has been losing share to ChatGPT, Perplexity, and Claude since 2024. AI Overviews cannibalize Google's own clicks. Users are moving the starting point for their questions away from the search bar. In that context, "AEO and GEO are still SEO" is not a neutral observation. It is an attempt to keep the market on the terrain where Google is still strongest: ranking documents against keywords. An independent analysis would at least name Google's position and its incentives. This guide does not, and most of the articles repeating it do not either.
What a guide about search cannot decide
Google can advise on what works inside its own index. Google cannot decide whether ChatGPT, Perplexity, or Claude cite you in their answers. Those systems draw on their own training data and their own live web lookups, with their own criteria for what counts as a credible source. Advice that says "ignore llms.txt" says something about how Google sees the world, not about how other AI systems read it. That difference decides who gets cited six months from now and who does not.
What the data actually shows
Ahrefs compared 15,000 queries across four AI assistants and Google. 80% of the sources AI cites do not even rank in Google's top 100. That is not overlap. That is a different market.
In August 2025, Ahrefs published an analysis of 15,000 long-tail queries comparing the URLs cited by ChatGPT, Gemini, Copilot, and Perplexity against Google's search results for the same queries. The result made the entire "AI search is just SEO" frame untenable. Only twelve percent of AI-cited URLs appeared in Google's top ten for the same prompt. Twenty percent appeared anywhere in the top hundred. Eighty percent did not appear at all. If most of what AI systems cite does not even rank in Google's top hundred, then AI is not drawing from the same pool Google is. "It is still SEO" is not rhetorically wrong. It is empirically wrong.
PageRank and EntityRank are two different games
PageRank ranks documents against a query. AI systems recognize entities and call them up inside an answer. The same expert can win one game and lose the other. The Ahrefs numbers prove the games rarely match.
PageRank, the algorithm Google launched with in 1998, counted links. A page that other pages linked to was probably important. Twenty-five years later, in 2026, large language models do something different. They compose answers by drawing on an internal picture of who and what exists in the world. The entities a model recognizes, and trusts the connections of, end up in its answers. The ones it does not recognize do not. We call that EntityRank internally. It is not an official term. No engineer at OpenAI or Google uses it. But it describes what is happening more accurately than "SEO 2.0." A strong entity does not need a top-ten ranking to be cited. A top-three ranking does not guarantee citation. Sometimes the two align, often they do not. That is the whole point.
What an AI system actually values under the hood
Follower counts do not feed EntityRank. Keyword density does not feed it. High publishing volume does not feed it. What does feed it: consistent naming across sites, platforms, podcasts, and articles. External mentions in places that themselves carry authority. Topic clusters that show depth on one subject instead of scattered pages about everything. Structured data that communicates entity relationships, not AI-specific syntax. Those are different signals than a 2018 SEO checklist.
What Google gets right, and what it does not
Google is right that fundamentals matter. Google is wrong that fundamentals only count through Google's lens. AI traffic is growing faster and converting five times better. Building only for Google means missing the market that is rising now.
Google's guide has one point in its favor: content quality and authority signals still matter. A weak page dressed up for AI crawlers does not beat a strong, substantive page from a recognized expert. What Google gets wrong is the assumption that quality and authority can only be measured through Google's lens. An expert who is consistently cited by ChatGPT without ever reaching Google's top ten proves that visibility itself has fragmented. Two systems, two judges, two outcomes. Anyone optimizing for one of them is missing the other. And the other one is growing faster. AI referral clicks grew 527% year over year. Conversion rates on AI-referred traffic sit around 14%, compared to roughly 3% from Google organic. That is not a shift coming in the future. It is already happening.
What this means for experts who want to be found in 2026
Not ranking harder, but becoming recognizable as an entity. That demands identity infrastructure, not more content volume. The experts building that layer now are the ones AI will recommend six months from now.
The practical task is not to rank harder. It is to become recognizable as an entity. That starts with one clear identity and positioning, expressed consistently across your own domain, podcasts, articles, interviews, and every place AI systems read the web. Building topic clusters around one subject instead of scattered content about everything. Earning external mentions in places that themselves carry weight. Using structured data to make entity relationships legible to systems that read context, not systems that count keywords. Research Identity First Media relies on shows a prospective client needs to consume between two and seven hours of content before there is enough trust to buy. In an AI-mediated discovery environment, that content has to exist, and it has to be traceable back to you as an entity the systems recognize.
The question most experts skip
The easy question is: how do I rank higher on Google? The harder question is: does the entity I am trying to be exist clearly enough that AI systems can call me up when someone asks a question I am the answer to? That is not a content-volume problem. It is an identity-infrastructure problem. And it is precisely the layer Identity First Media is built on: a platform that turns your identity, expertise, and consistent output into input that AI systems can read, recognize, and cite.
Frequently Asked Questions
Did Google officially declare that AEO and GEO do not exist?
Google's new guide says AEO and GEO are extensions of SEO, not separate disciplines. But Ahrefs research across 15,000 queries shows 80% of the URLs AI systems cite do not even appear in Google's top 100. Empirically, Google's framing does not hold up. It is a market statement from a player losing share, not an objective technical analysis.
Should I add an llms.txt file to my site or not?
Google says to skip it. But Google does not decide whether ChatGPT, Perplexity, or Claude cite your site in their answers. Those systems have their own logic. An llms.txt file is a lightweight signal for AI crawlers and costs almost nothing to set up. The bigger question is whether your entity architecture is in order: consistent naming, topic clusters, external mentions, structured data. That decides whether you get cited, not a single text file.
What is the difference between ranking in Google and being cited by AI?
Google ranks documents against keywords for a specific query. AI systems recognize entities and call them up inside an answer, based on an internal picture of who carries authority on a theme. According to Ahrefs, 80% of the URLs AI cites do not appear in Google's top 100 for the same query. Those are two different mechanics that sometimes overlap, often do not. A strong entity can be cited without ranking high. A top-ten ranking is no guarantee of citation.
Why would Google publish a guide like this now?
Google has been losing share to AI assistants since 2024. AI Overviews cannibalize Google's own clicks. Users are moving the starting point for their questions to ChatGPT, Perplexity, and Claude. A guide that says "AEO and GEO are still SEO" keeps the market on terrain where Google is strongest: ranking documents against keywords. That is not an objective analysis. It is a commercial move from a platform trying to reclaim ground.
What should an expert actually do to show up in AI answers?
Work on recognition as an entity, not on ranking. One clear identity, expressed consistently across your own domain, podcasts, articles, and external mentions. Topic clusters built around one subject. Structured data that makes entity relationships visible. Enough deep content that a prospective client can build the picture they need within two to seven hours of consumption. That is the Identity-First Methodology in practice, and it is exactly what AI systems pick up on.
Discover in 2 minutes how visible you are to AI like ChatGPT, Claude and Gemini.
Start your free scan