By Eric Richmond, Founder and CPO, CiteMetrix
Six days ago, Google’s Search Central team published a new official guide called “Optimizing your website for generative AI features on Google Search.” Within hours, every SEO newsletter on LinkedIn was quoting the same sentence:
“From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”
It was framed as Google ending the generative engine optimization (GEO) debate. The message? Just do SEO. There’s nothing new to learn. There’s nothing new to buy.
I read the post carefully, twice. There is a lot in it that is genuinely true, and the SEO community is right to take the technical advice seriously. But there is also a structural problem with the entire piece that nobody seems to be naming, and it matters enormously to anyone whose job is actually protecting a brand in the AI-search era.
Here is the part Google got right, the part Google didn’t get wrong but also didn’t say, and the part that should have been in the guide and wasn’t.
What Google got right
If you’re running a website in 2026, you should still do classic SEO. Crawlable HTML. Clean technical structure. Non-commodity content written by humans with actual expertise. Good page experience. Structured data where it earns its keep.
None of that has changed, and the post is correct that Google’s AI Overviews and AI Mode are grounded in Google’s existing ranking index. The features are downstream of the index. If you’re invisible in the index, you’re invisible in the AI surface that sits on top of it.
The myth-busting section is also useful as a corrective. Yes, you can ignore the cottage industry of “GEO hacks” that promise to game AI rankings through special llms.txt files, paragraph chunking, or rewriting your H1s to address the AI directly. Google says these tactics don’t work for Google Search, and on Google Search, they’re almost certainly right.
The post is also right to flag the rise of agentic browsing. AI agents are starting to access websites the way a screen reader does, analyzing the rendered DOM, inspecting the accessibility tree, and executing transactions on behalf of a user. If your site is hard for an assistive technology to parse, it’s going to be hard for an agent to parse. That is a real technical concern, and it deserves the prominence the post gives it.
So far, so reasonable.
What Google quietly omitted
Here is the sentence in the post that bears almost the entire weight of its argument. I want to put it in front of you again, in the original context, so you can see what it actually says:
“From Google Search’s perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO.”
Read that phrase one more time: From Google Search’s perspective.
The post is, by its own framing, a guide to optimizing for Google. It is not a guide to AI search visibility as a whole. It is a guide to optimizing for Google’s AI features, AI Overviews, AI Mode, and the rest of Google’s surfaces.
Eight of the nine AI platforms that brands now need to think about, ChatGPT, Claude, Perplexity, Copilot, Mistral, Grok, DeepSeek, and Google’s own Gemini standalone product (which is different from AI Overviews), are not mentioned in the post a single time. The word “Perplexity” does not appear. The word “ChatGPT” does not appear. The word “Claude” does not appear.
The Google guide is silent on whether anything it recommends generalizes to the rest of the AI ecosystem because that ecosystem is, structurally, not Google’s problem to solve.
This is not a criticism of Google. Google is allowed to publish a guide to its own surfaces. What I am criticizing is the way the post has been received, as if Google had ended the debate about how brands should approach the entire AI search landscape. It hasn’t. It has only spoken about one platform out of nine.
The structurally honest version of the post would have begun: “This guide describes how to be visible in Google’s AI features. To understand how to be visible across the wider AI search ecosystem, you will need to look elsewhere.”
The part Google could not have written
There is a deeper problem with the “GEO is just SEO” framing, and it has nothing to do with Google’s coverage of other platforms. It has to do with which buyer the post is written for.
Every word of Google’s guide is addressed to the SEO operator. The vocabulary is SEO vocabulary, crawlability, indexation, snippet eligibility, structured data, JavaScript rendering, robots.txt. The reader Google has in mind is the person who has been responsible for organic traffic for the last fifteen years and is now wondering whether AI changes their job.
For that person, Google’s answer is largely: no. If you do the technical basics well, you’ll likely appear in AI Overviews.
But the SEO operator is not the only person inside a company who cares about how that company shows up in AI answers. There is another buyer, with a completely different job to be done, and Google’s post is silent on their existence.
The brand manager: and the marketing agency that serves them: has a different set of questions. They aren’t just asking “how do I rank?” They are asking:
1. How is my brand being described in AI answers right now?
Not whether the brand ranks, but whether the AI tells the customer the truth. A regional hospital network might rank #1 for its own name on Google and still have ChatGPT confidently telling customers its emergency room is closed on weekends because of an outdated forum post. AI brand monitoring catches that; traditional SEO does not.
2. How am I doing against my competitors in AI answers?
When a customer asks Perplexity for the “best wealth management firm in Hartford,” am I in that answer set? Where do I sit compared to my three closest competitors? This is a share-of-voice question, and you cannot answer it by looking at Google Search Console.
3. Is my answer accurate enough to defend in front of a compliance committee?
In healthcare, finance, or legal, AI hallucinations aren’t just bad marketing: they are regulatory risks. If an AI makes a false claim about your insurance coverage or medical advice, who is responsible? Google’s guide doesn’t mention the word hallucination once.
These are not SEO questions. They have nothing to do with crawlability. They are measurement and reputation questions across nine different platforms.
Why the framing matters
Google’s post will likely be cited for years by consultants who want to tell clients there is “nothing new to see here.” The danger is that brand managers will believe it, only to be blindsided when a journalist or investor surfaces an AI answer about their company that is wrong, damaging, or completely favors a competitor.
That blindside is exactly what CiteMetrix was built to prevent.
We aren’t an SEO tool, and we don’t compete with the SEO stack. Instead, we monitor how nine major AI platforms describe your brand. We score the result with a single composite metric called ModelScore™, detect hallucinations in real-time, and let you see how your share of voice compares to the competition.
The buyer isn’t just the person looking at crawl logs; it’s the person responsible for the brand’s perception in the world. And increasingly, that perception is being formed inside a chat window.
Conclusion: Use the right tool for the right job
If you are doing SEO work, follow Google’s guide. The advice is technically sound for Google’s specific surfaces.
But if you are responsible for your brand’s accuracy, sentiment, and competitive positioning across the entire AI search ecosystem, you need more than a robots.txt audit. You need a dedicated strategy for ai search visibility.
Eight of the nine platforms you need to think about aren’t in Google’s guide. Hallucinations aren’t in it. Sentiment isn’t in it. Share of voice isn’t in it.
That gap isn’t a bug in Google’s documentation: it’s the reason CiteMetrix exists.
