What it is
Google published its position on GEO on 15 May — nearly two months ago.
That day Search Central posted an announcement and shipped a new document, Optimizing your website for generative AI features on Google Search, under a section created to hold it, covering “mythbusting common AEO/GEO misconceptions” by name. It opens with: “From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.” The document was updated again on 10 July.
Two more followed on 5 June: a brand-new page on third-party SEO tools and services, and a new section in the long-standing “Do you need an SEO?” guide on how to vet a vendor — with “AEO” and “GEO” services named explicitly.
One more thing landed in the same window. On 7 July, Search Console added platform properties, so people with no website of their own — running only Instagram, TikTok, X or YouTube — can now see how their content performs in Google Search and Discover. For brands, that means your content on third-party platforms finally has an official visibility number attached to it. (This is separate from June’s generative AI performance report: that one covers your site’s impressions inside AI features, this one covers how your social accounts perform in Search.)
What Google crossed off is exactly what the industry is selling
The AI guide lists what you don’t need to do: you don’t need an llms.txt or any AI-specific file to be understood; you don’t need to break your content into tiny pieces for models to digest; you don’t need to write in some special register for generative search; and chasing “mentions” around the web isn’t as helpful as it looks.
Every one of those maps onto something being packaged and sold right now. llms.txt is the clearest case — no major engine has ever announced support for it, yet this year it got talked about as infrastructure. Google has now crossed it off its own list.
The guide also says structured data isn’t a requirement, and that line has to be read within its own scope: Google means it isn’t required for Google’s generative features, because its ranking systems already recognize you through other signals. For engines without that stack underneath them, machine-readable markup remains one of the few handles they have.
But “GEO is still SEO” opens with five words
“From Google Search’s perspective.”
Those five words are Google’s own, and they fence the document’s scope precisely: Google is talking about AI Overviews and AI Mode, features that live inside Google Search. By its account the machinery underneath is the same, so it counts as SEO.
Your customers aren’t only asking their questions there. ChatGPT, Perplexity and Copilot each carry their own index and their own citation logic. Google has no jurisdiction over them and no obligation to speak for them. Carry “GEO is still SEO” outside Google’s own gate and Google hasn’t endorsed it either.
What’s happening in the market right now is a wave of people quoting that line with the first five words cut off, as a conclusion about the whole of AI search.
Even inside Google’s own house, the line isn’t as useful as it sounds. Across multiple rounds of our own testing, pages built strictly to Google’s SEO advice had a relatively hard time getting into AIO; under the same conditions, our GEO approach performed several times better. The guide makes “do SEO well and AI will find you” sound smooth. How wide the gap underneath it runs, you only learn by testing.
It also handed clients a checklist for vetting vendors
The third-party page is blunt. Its words: “Google doesn’t evaluate third-party services, so be wary of such claims and those making them.” And: “Third-party tools don’t have access to our internal ranking data. They can’t guarantee performance.” Plus an observation about the tooling industry as a whole — some third-party services hand you data that users misread as coming from Google.
The three companion questions sit in the updated “Do you need an SEO?” guide:
- Do they cite official Google documentation as supporting evidence for their recommendations?
- If they have advice on optimizing for AI experiences (also known as “AEO” or “GEO” services), is their advice aligned with Google Search’s official guidance on optimizing for generative AI features?
- Do they use tools that are aligned with Google’s guidance?
Keep that checklist. It asks sharper questions than most buyers think to ask. The ones it will trip up are the ones printing “guaranteed AI citations” on a price list.
So we’re keeping the word “GEO”
Google says AI Overviews and AI Mode are rooted in its core ranking systems. As a description of the mechanism that may hold to some degree — it’s the only party who can see inside its own systems.
What doesn’t hold is the sentence it draws from that mechanism: do SEO well and the generative features will find you. Underneath one system, the conditions for making ten blue links and the conditions for being pulled into a generated answer are not the same set. That is the gap those numbers measure — and it shows up without ever leaving Google’s own turf.
And the equals sign has been carried out of Google’s gate on top of that. For the engines it has no reach into, its guide says nothing at all, yet people quote it as a conclusion about all of them.
So the word stays. “GEO” marks the half Google won’t answer for you — the half with no official documentation, no console to open, and no one guaranteeing performance, the half you can only chase by testing. Folding it back into SEO amounts to declaring that half doesn’t exist.
Should you act now
Three things, in order.
One, read the AI guide yourself. It’s short, and it’s the primary source for all of this; you’ve seen plenty of second-hand versions over the past two months.
Two, put those three questions to your current vendor verbatim. No answer, or an answer along the lines of “that’s our proprietary algorithm,” and you’re looking at exactly what the document was written to protect you from.
Three, the half Google can’t reach has to be measured separately. Who gets cited inside ChatGPT or Perplexity, and which version of you they cite, will never appear anywhere in Google’s documentation. There’s no official console to open for that half — it takes sustained, cross-engine, repeated testing, because the same question asked today and asked next week will name someone else. The tedious part is that it has no finish line, and no amount of guide-reading makes it easier. That’s the part we watch for clients (see what managed covers).
Do the first one. You’ll notice it only ever talks about its own half — and has no intention of answering for the rest.