TL;DR — Google’s 15 May 2026 guidance says GEO and AEO are made-up terms, and that optimising for AI search inside Google is just SEO. They’re correct, but only for Google’s own AI products. ChatGPT, Perplexity, Claude and the rest run separate systems with separate rules. Here’s what actually moves the needle, and the four “myths” Google called out that aren’t as settled as they look.
On 15 May 2026, Google published its first official guidance on optimising for AI search. It’s a useful document. It’s also a quietly self-interested one, and the difference between those two things matters if you run a business that depends on being found online.
The headline takeaway, in Google’s own words:
“From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”
In other words: GEO and AEO are made-up terms. Stop worrying about them. Do good SEO and you’ll show up in AI Overviews too.
Key Takeaways
- Is GEO a real thing? What Google actually said
- What Google quietly left out
- Should you create an llms.txt file? (And the other three “myths”)
- The strategic question this raises
Google then went further and named four specific tactics they say you can ignore — llms.txt files, “chunking” your content for AI, rewriting pages in an AI-friendly style, and chasing brand mentions. The SEO world has spent the last week arguing about it.
I’ve read the guidance carefully. Here’s what I think it actually means, and where I think Google is being economical with the full picture.
Is GEO a real thing? What Google actually said
Short answer: as a separate discipline from SEO, Google says no. As a useful frame for thinking about AI search visibility outside Google’s own products, yes. The truth depends on which AI system you care about being cited by.
If you only care about being cited inside Google’s AI products — AI Overviews, AI Mode, the things that appear on a google.com results page — then yes, it’s still SEO. The systems that power those features pull from the same index, ranked by the same algorithms, weighted by the same quality signals. A site that ranks well in regular Google Search is the same site that gets pulled into an AI Overview.
The technical fundamentals Google lists are correct and worth repeating: be crawlable, be indexable, write for humans, structure your content clearly, have real expertise on the page, make the site work on phones. None of that is new. None of it is controversial. None of it will be obsolete in five years.
The bit about chasing inauthentic mentions is also worth taking seriously. Paying for fake citations, manufacturing PR, getting your name dropped on low-quality blog networks — none of it works, and Google’s systems are good at spotting it. What works is the boring version: real customer reviews, real industry write-ups, real expertise that other people choose to reference because it’s actually useful.
So far, so reasonable.
What Google quietly left out
Short answer: Google’s guidance is about Google’s AI products. It doesn’t cover ChatGPT, Perplexity, Claude or the other AI tools your customers are increasingly using to find businesses — and each of those systems plays by different rules.
When a customer types a question into ChatGPT — “best physiotherapist in Bristol,” “is acupuncture worth trying for sciatica,” “who runs wedding photography in Yorkshire” — Google has no involvement in the answer. ChatGPT doesn’t query Google. It runs its own retrieval system, with its own crawler, its own ranking signals, its own rules about what makes a source citable.
Same goes for Perplexity. Same for Claude. Same for the dozen other AI tools your customers are already using to find businesses.
Each of these systems has independent answers to questions like:
— Which crawlers can reach my site?
— Do I read llms.txt?
— Do I weight schema markup heavily?
— How do I decide which sources to cite when I generate an answer?
The systems disagree with each other. There is no single playbook. Anyone telling you otherwise is selling something.
So when Google says “optimising for AI search is just SEO” — they mean Google’s AI search. The sentence is accurate within its scope. The scope is much smaller than the sentence implies.
Should you create an llms.txt file? (And the other three “myths”)
Short answer: Google doesn’t use llms.txt, so for ranking in AI Overviews it makes no difference. But llms.txt was never built for Google — it’s a draft standard for the wider AI ecosystem, and some non-Google crawlers do respect it. Whether it’s worth your time depends on which AI systems matter to your business.
Worth going through all four myths properly, because the truth is more useful than the slogan.
On llms.txt. Google doesn’t use it. They never have, and based on the new guidance, they’re not going to. That’s fine — llms.txt was never proposed for Google. It’s a draft standard for the wider AI ecosystem, and adoption is patchy. Some crawlers respect it, some ignore it, some are still deciding. Dismissing it because Google specifically doesn’t use it is like dismissing a phone charger because it doesn’t fit one brand of laptop. Different systems, different rules.
On chunking content. Google’s right that you shouldn’t fragment pages artificially. But there’s a real signal underneath the bad advice, which is that AI systems extract short, self-contained passages when they cite you. A page where the answer is buried in a 600-word paragraph gets cited less often than a page where the answer is plainly stated in the first three sentences. That’s not chunking. That’s just clearer writing. Most business websites would benefit from it whether AI existed or not.
On rewriting content for AI. Same shape of argument. You don’t need a special AI register. You do need to answer the actual question a customer would ask, in plain language, near the top of the page. If your homepage’s first paragraph doesn’t say what you do and where you do it, no AI is going to cite you, and no human is going to scroll. Writing clearly is not an AI hack. It’s just writing.
On inauthentic mentions. Google’s entirely right on this one. Don’t pay for fake citations. Don’t run link schemes. Don’t try to game brand mention counts. Do the work that earns real recognition — and accept that it’s slower than the alternative.
The strategic question this raises
Here’s the more interesting question, and the one I don’t see being asked much.
Google has just publicly told the SEO industry that the rules for AI search are the same as the rules for regular search. That message benefits Google enormously. It positions Google’s index as the centre of gravity for AI search at exactly the moment when ChatGPT, Perplexity and Claude are eating into Google’s traffic share. It tells businesses they don’t need to think about anyone else’s system. It quietly anchors the conversation back to Google.
I’m not accusing Google of being dishonest. The guidance is fundamentally accurate for the part of the AI search world Google controls. But I would be cautious about taking strategic advice on the entire AI search landscape from the player with the most to lose if that landscape fragments away from them.
If your customers are using ChatGPT to find businesses like yours — and an increasing share are — then “do good Google SEO” is necessary but not sufficient.
A five-step plan for AI search visibility in 2026
If you run a small business and you’re trying to work out where to put your effort, this is the order I’d think about it in.
1. Get the SEO fundamentals right. Google’s guidance on this is correct and free. A site that ranks well in Google is a site that’s already most of the way to being AI-visible inside Google’s products. Don’t skip this step looking for clever shortcuts.
2. Make your content extractable. Every important answer about your business — what you do, where, for whom, what it costs, why someone should pick you — should be statable in two or three plain sentences that an AI can quote. Most business sites fail this test badly.
3. Check whether the AI crawlers can actually reach your site. ChatGPT, Perplexity and the others use specific user-agents. They get blocked by default on a surprising number of hosts. If they can’t crawl you, they can’t cite you, and no amount of SEO work will change that.
4. Build the citations that compound. Real reviews, real press, real industry mentions. Slow work. Worth it.
5. Measure. You can’t fix what you can’t see, and AI visibility is invisible by default — there’s no Search Console for ChatGPT. This is where most businesses get stuck, and it’s the gap that tools like SearchScore exist to close. (Full disclosure: I built it.) Whatever tool you use, the principle stands: pick one, run it monthly, and treat the score as a baseline rather than a verdict.
Where this leaves us
Google’s guidance is genuinely useful for the part of the AI search world it covers. It’s also a reminder that the people setting the rules are not neutral parties, and “official guidance” from any single platform is one input, not the whole picture.
The businesses that win in AI search over the next few years will be the ones that take Google’s fundamentals seriously and recognise that the fundamentals don’t end at Google’s edge. Both things are true. Both matter. Don’t let anyone — including Google — convince you it’s only one of them.
Frequently asked questions
Is GEO different from SEO?
Google’s official position is no — they say optimising for their AI features is just SEO. But for AI systems outside Google (ChatGPT, Perplexity, Claude), the signals that drive citation are different enough that treating “GEO” as a distinct practice is useful. The honest answer depends on which AI system you want to be visible in.
Does Google use llms.txt?
No. Google’s May 2026 guidance confirmed they don’t read llms.txt files and creating one gives no advantage in Google AI Overviews or AI Mode. However, llms.txt is a proposed standard for the wider AI ecosystem — some non-Google crawlers do respect it, though adoption is still inconsistent.
How do I get cited by ChatGPT?
Make sure ChatGPT’s crawler (GPTBot and OAI-SearchBot) can access your site, structure your content so answers are stated plainly in short paragraphs near the top of each page, add accurate schema markup, and build genuine third-party citations. There’s no single setting that guarantees citation — it’s the same fundamentals applied with AI extractability in mind.
Should I rewrite my content for AI search?
No, not in a special AI register. But you should make sure the answer to common customer questions is stated in plain, self-contained sentences near the top of each page. That helps AI systems quote you and helps human readers find what they need faster.
What is Google AI Mode?
AI Mode is Google’s conversational AI search experience, separate from the AI Overviews that appear above traditional search results. Both are powered by Google’s core ranking systems, which is why Google’s guidance treats optimisation for them as standard SEO work.
