- A viral thread from SEO consultant Charles Floate argues that Google’s public AI search guidance downplays the role of chunks, source quality and machine-readable content structures.
- Google says publishers do not need special AI files, chunked pages or AI-specific rewrites for generative AI search, but the debate shows how unsettled AI SEO still is.
A viral thread from SEO consultant Charles Floate has reignited one of the biggest arguments in search right now: is AI search still just SEO, or are the systems behind AI answers changing the rules more than Google admits?
The thread reacts to Google’s recent guidance on generative AI search, where the company pushed back against several popular AI SEO tactics. As The Query Post reported last week, Google’s position is that AI search optimization is still SEO, not a separate discipline built around special files, artificial chunking, AI-only rewrites, or new markup. But as we wrote in a follow-up analysis, many SEOs are not fully convinced that Google’s public guidance captures everything changing under the surface of AI-powered search.
ANTHROPIC IS TARGETING SEO CONTENT! 🚨
AND GOOGLE USES CHUNKS, MARKDOWN FILES AND EVEN LLMS.TXT IN IT’S OWN INTERNAL SYSTEMS!!! 👀
Meanwhile Google’s own AI SEO guide tells you none of this even matters…
Anthropic literally wrote into Claude’s system prompt that SEO content… https://t.co/sihP0g843L pic.twitter.com/Q4NRtPHbzD
— Charles Floate 📈 (@Charles_SEO) May 18, 2026
Floate argues that this public guidance is incomplete. His claim is that modern AI systems still rely heavily on retrieval, source-quality filtering, extracted content fragments and internal instructions that may shape which sources are trusted or ignored.
The Dispute Starts With Google’s AI SEO Guidance
Google’s message is straightforward: keep making useful, crawlable, people-first content. The company is trying to discourage publishers from chasing every new AI SEO hack, especially tactics like creating special machine-readable files, rewriting content only for AI systems, or manufacturing mentions across the web.
That position is also consistent with Google’s broader spam direction. The Verge reported that Google updated its spam rules to include attempts to manipulate generative AI responses in Search, including AI Overviews and AI Mode.
For Google, the message is defensive as much as instructional. The company does not want publishers building pages for AI systems in a way that turns generative search into another easily gamed SEO surface.
Floate’s Counterargument: Retrieval Still Changes the Game
Floate’s argument is that Google’s guidance may be technically true in a narrow sense but still misleading for operators. His thread says the important layer is not whether Google officially requires “chunking” or llms.txt, but whether AI systems retrieve, filter and synthesize information in smaller, source-referenced units.
In his view, AI visibility depends less on the full page as a finished article and more on whether specific sections can be extracted, understood and trusted as useful answer material.
That is a more aggressive interpretation than Google’s public guidance, and some of it depends on claims around leaked or reverse-engineered system prompts. Those claims should be treated carefully, because private system prompts are difficult to verify independently. Still, the broader retrieval argument is not fringe.
A recent academic study on Google Search, Gemini and AI Overviews found that AI Overviews and Gemini retrieve and present sources differently from traditional Google Search. The study reported that AI Overviews appeared for 51.5% of representative real-user queries in its benchmark and that source overlap between traditional search and generative search systems was low.
The Chunking Debate Is Really About Extractability
One of the most contested points is chunking. Google says publishers do not need to break content into tiny pieces for AI systems to understand it. Floate argues that this misses the practical reality of retrieval systems, where answers are often assembled from snippets, passages or source fragments rather than full-document reading.
The safest way to frame this is simple: publishers probably do not need to artificially chop pages into tiny blocks for Google. But clear section structure still matters.
Pages with descriptive headings, self-contained sections, clean lists, direct answers and readable HTML are easier for both users and retrieval systems to parse. That is not necessarily a new SEO discipline, but it does make “write clearly and structure well” more important in an AI search environment.
The Mention Question Is Also More Complicated
Google’s guide warns against seeking inauthentic mentions across the web. Floate agrees with the anti-spam part, but argues that high-quality third-party mentions still matter because AI systems may use source quality and corroboration when deciding what to trust.
That distinction matters. A paid mention on a low-quality listicle is not the same as being cited by a primary source, a respected industry publication, a government site, a research paper, or a credible company blog.
Research into conversational and generative search also suggests that manipulating AI visibility is harder than some GEO vendors suggest. A 2025 benchmark paper on Conversational SEO found that many proposed C-SEO methods were less effective across broad testing than earlier claims suggested, while traditional SEO signals that help a source appear in the model’s context remained more important.
That supports a more balanced view: artificial mention-building may be risky and ineffective, but earning real mentions from trusted sources still looks strategically important.
Google May Be Right and Still Not Say Everything
The most useful takeaway is not that Google is “lying.” It is that Google’s public guidance and the operational reality of AI retrieval may answer different questions.
Google is telling publishers what not to over-optimize for: special files, artificial chunking, AI-only rewrites, spammy mentions and unnecessary schema. That advice is useful.
But AI search systems still need to retrieve, evaluate and cite sources. That means clarity, authority, source reputation, technical accessibility and structured sections can still matter, even if Google does not call those things “GEO.”
What SEOs Should Actually Do
The practical answer is not to chase every AI SEO trick. It is also not to ignore AI search completely.
For publishers and marketers, the safest approach is:
- Make pages crawlable and indexable.
- Use clear headings and self-contained sections.
- Answer important questions directly.
- Support claims with primary sources where possible.
- Build real authority through credible mentions, not artificial citation schemes.
- Avoid creating AI-only doorway content or manipulative recommendation pages.
That may sound like traditional SEO, and in many ways it is. But AI search raises the stakes for clarity and source trust because the user may never click through to inspect the full page.
The Query Post View
The viral thread shows why the SEO industry is still struggling to agree on what AI search actually changes.
Google’s public message is clear: AEO and GEO are not separate disciplines. From Google’s perspective, optimizing for AI search still means doing SEO well. That means useful content, crawlable pages, clear structure, strong sources, and no artificial tricks built only for generative systems.
But the pushback from SEOs is not coming from nowhere. AI search does not behave exactly like the old ten-blue-links model. It retrieves information, compresses it, cites some sources, ignores others, and turns pages into answer material before a user ever sees a search result. That changes the visibility game, even if the fundamentals still overlap with SEO.
The mistake is treating this as an either-or debate.
Google is probably right that there is no magic AI SEO playbook built around llms.txt files, artificial chunking, fake mentions, or special markup. But SEOs are also right to question whether “just keep doing SEO” fully explains how AI-generated answers choose, summarize, and cite sources.
The practical middle ground is where this gets interesting. AI search does not reward tricks for long, but it does reward content that is easy to retrieve, easy to understand, and easy to trust. That means clearer sections, stronger sourcing, more original information, better entity signals, and a web presence that looks credible beyond one optimized page.
So The Query Post’s view is simple: GEO may be overhyped as a separate industry label, but AI visibility is real. The best strategy is not to chase every new acronym. It is to build pages and sources that can survive both classic search and AI retrieval.
That means writing for humans first, but structuring for machines enough that they can actually extract the value. In 2026, that may be the real SEO skill: not gaming AI systems, but becoming the kind of source they can confidently use without having to guess.
