Google Just Said AEO Is Just SEO. Two Caveats Worth Knowing.

Google Search Central published its official guide on optimising for generative AI features in June 2026. It is worth reading carefully, because it does two useful things at once. It validates the SEO foundations most reputable practitioners have been recommending for years. And it explicitly throws out a long list of "AI search hacks" that are now circulating as essential. Both are useful corrections.

This piece walks through what Google actually said, where we agree with it, and the two caveats that matter for UK service SMEs trying to be visible in generative AI search.

What Google actually said

The headline message is clean. Google's generative AI features (AI Overviews and AI Mode) are rooted in the same core Search ranking and quality systems that power traditional search. They use retrieval-augmented generation (RAG) to fetch pages from the Search index and query fan-out to surface related results. From Google's perspective, "optimising for generative AI search is optimising for the search experience, and thus still SEO."

The guide then lists what Google says actually works:

• Create unique, valuable, non-commodity content with a clear point of view
• Maintain a clear technical structure (meet Search technical requirements, follow crawling best practices, sensible semantic HTML, JavaScript SEO basics, good page experience)
• Reduce duplicate content
• Use structured data as part of overall SEO (still useful for rich results, not required for generative AI specifically)
• For local and ecommerce, keep Google Business Profiles and Merchant Center feeds current
• Explore agentic experiences (browser agents, emerging protocols like UCP) as a forward-looking item

And here is the bit getting most of the attention. Google's "mythbusting" section explicitly tells you what does not work:

• LLMS.txt files and other "special" AI markup. Google Search does not use them.
• Chunking content into tiny pieces. Not required. Google understands multi-topic pages.
• Rewriting content just for AI systems. AI understands synonyms and meanings.
• Seeking inauthentic mentions across the web. Not as helpful as it seems.
• Overfocusing on structured data. Not required for generative AI, though still useful for rich results.

That is a meaningful list. Each of those items has been sold as an AEO necessity over the last twelve months. Google is telling you, on the record, that you can stop worrying about them.

Where we agree with Google

Almost completely. The Optimus AI Visibility audit has always been built on a small number of things that map directly onto Google's foundations:

• Unique, expert-led content with a clear point of view
• Clean technical foundations (crawl, index, page experience, structured data where it earns its keep)
• A clear site architecture that helps humans first and search engines second
• Authority signals built through genuine expertise, not bought mentions
• Active monitoring through Search Console so you know what is actually happening

None of those are AEO hacks. All of them are the foundations Google's guide endorses. The audits we have been running for UK service SMEs, the AEO scores we have been producing, and the action lists we have been writing all sit comfortably inside Google's framework. We have not been chasing LLMS.txt files, recommending content chunking, or selling inauthentic-mention packages. The "AI Visibility" service is essentially "are your foundations good enough that the AI systems can actually use you", and that aligns directly with how Google now describes the work.

The two caveats Google's guide leaves out

This is where the piece earns its keep. The guide is excellent for what it is, but it is a Google document about Google Search. Two things sit outside its scope and matter for UK service SMEs in 2026.

Caveat one: Google is one search engine. ChatGPT, Claude, Perplexity and others do not all behave the same way.

When a UK consumer asks an AI assistant whether to switch their insurance, the answer might come from ChatGPT, from Perplexity, from the Aviva ChatGPT app, from a Claude integration inside a browser, or from a Copilot embedded in Microsoft 365. Each of these uses different retrieval logic. Some lean heavily on Bing's index. Some lean on partner data feeds. Some use their own crawl. Some use a hybrid approach.

Google's guide describes how Google Search's AI features work. It does not, and cannot, describe how every other AI engine works. A UK service SME trying to be visible across "AI search" needs to think about the multi-engine reality, not just the Google one. That is the gap a proper AI Visibility audit fills. The foundations Google describes are necessary. They are not always sufficient on the other engines, where things like clean machine-readable structured data, authority signals across third-party platforms, and yes, sometimes specific formats like FAQ schema, still do meaningful work.

Caveat two: "Doing SEO well" still requires understanding how generative AI actually retrieves and presents your content.

Google's guide makes the point that the fundamentals are the work. True. But the work has to be done with an understanding of how generative AI actually uses content. The same page can score well in classic SEO and badly in generative AI retrieval if it is structured for human scanning rather than for AI extraction. A page where the answer to a likely user question sits in paragraph three rather than paragraph one is still good SEO. It may or may not be the version Google's RAG layer pulls cleanly into an AI Overview.

That is not a hack. It is a content design discipline. And it is one of the things a good AI visibility audit surfaces that a traditional SEO audit might not.

What this means for UK service SMEs

Three things, practically.

First, if anyone is selling you LLMS.txt files, content chunking services, or inauthentic-mention packages as essential AI visibility tactics, Google has now told you on the record that they are not. You can decline politely and save the money.

Second, the SEO foundations Google describes are not optional. Unique, expert-led content. Clean technical structure. Good page experience. Honest authority signals. If you have not invested in those in the last two years, that is the conversation worth having before any AI Visibility one.

Third, the multi-engine reality still matters. ChatGPT, Claude and Perplexity are not Google, and being visible on each one needs to be checked separately. That is the bit Google's guide cannot help with by design, and the bit a UK service SME needs to take seriously if AI distribution becomes a material channel in the next twelve months (which, on current evidence, it will).

A practical closing

Google's guide is a useful reset. It strips out the noise and puts the foundations back at the centre of the conversation. We agree with almost all of it. The two caveats are about what the guide cannot cover by design, not where Google is wrong.

If you have read this far and want to know where your own foundations actually sit, the Optimus AI Visibility audit is the easiest way to find out. We do not chase hacks. We check the foundations across the engines that matter, score what is there, and write the practical fix list.