Somewhere between a cottage industry selling llms.txt audits and a much smaller group of people who actually read Google's documentation, Google decided to just answer the question itself.
What Google actually published
In May 2026, Google Search Central published "Optimizing your website for generative AI features on Google Search," a guide covering AI Overviews and AI Mode specifically. It reframes the entire AEO/GEO conversation in one line: "From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO." That's not a subtle jab. It's Google declining to treat generative engine optimization as a discipline separate from the SEO fundamentals it has published guidance on for two decades.
The timing isn't incidental. Gary Illyes and Cherry Prommawin had already previewed pieces of this position at Search Central Live events, and Danny Sullivan had discouraged content-chunking tactics as early as January 2026. The May guide consolidates scattered public statements into one canonical document, which is exactly what turns "a Googler said something on a panel" into "cite this page." A conference remark is deniable. A permanent URL under developers.google.com is the thing every future disagreement with a vendor gets measured against.
It's also worth noticing what kind of document this is. Google didn't publish a blog post that will scroll off a feed in a week. It published fundamentals-tier documentation, the same tier as its crawling and indexing guides, its structured data reference, and its Core Web Vitals documentation. Google reserves that publishing tier for guidance it intends to stand behind for years, updated in place rather than superseded by a new post. Treating a fundamentals-tier page as a temporary statement is a mistake; treating it as the baseline every future GEO claim gets checked against is the more accurate read.
The five tactics Google says you don't need
The guide includes what amounts to a mythbusting section, naming specific tactics the GEO industry has sold hard over the past two years and stating plainly that none of them are required.
| TACTIC SOLD AS "GEO" | WHAT GOOGLE'S GUIDE ACTUALLY SAYS |
|---|---|
| llms.txt / AI-specific text files | "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search." |
| Content chunking | "There's no requirement to break your content into tiny pieces for AI to better understand it." |
| AI-specific rewriting | "You don't need to write in a specific way just for generative AI search." |
| Chasing inauthentic mentions | "Seeking inauthentic 'mentions' across the web isn't as helpful as it might seem." |
| Special AI schema markup | "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." |
“"You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search." (Google Search Central, Optimizing your website for generative AI features on Google Search)”
The llms.txt line lands hardest given how much of the GEO conversation this year has orbited that one file. We've tracked the data on llms.txt directly: two independent studies scanning a combined 337,894 domains found no citation bump from adopting it. And the contradiction runs deeper than one blog post. Chrome's Lighthouse tooling still flags llms.txt in its audits even as Google's own generative AI guide says it isn't needed for Search. Two parts of the same company, publishing guidance that doesn't fully agree with itself. That's not unusual for a company Google's size, but it's a useful reminder that a single signal from one Google surface isn't the same as an official ranking factor.
The other four items on the list get less attention than llms.txt, but they've been sold just as hard. Content chunking in particular has powered an entire genre of "restructure your content for AI" consulting over the past year, the pitch being that engines can only digest content broken into short, self-contained fragments. Google's guide addresses that directly: its systems already handle long, multi-topic pages, and forcing artificial fragmentation doesn't help retrieval. The inauthentic-mentions line is aimed at a specific cottage tactic too, seeding brand mentions across low-quality forums and comment sections purely to manufacture the appearance of third-party corroboration, which Google says isn't as helpful as vendors selling it would have you believe.
What Google says to do instead
The guide doesn't stop at debunking. It restates what it wants site owners doing, and none of it is new.
Notice what's absent from that list: nothing about schema types you haven't already deployed for standard SEO, nothing about a new sitemap format, nothing about rewriting your content in a different register for a different audience of one, the crawler. The guide's implicit argument is that a site already doing SEO fundamentals well doesn't need a separate GEO workstream bolted on top. It needs the fundamentals done more thoroughly.
The pointer to Search Console's Generative AI performance report deserves its own note, because it's a quiet correction to a whole category of third-party AI-visibility dashboards. Those tools estimate mention rate and citation share by running sample prompts against AI engines and inferring visibility from the results, a reasonable approach given they have no other option, but an estimate all the same. Google's own report, by contrast, is built from Search Console's actual impression and click data for AI surfaces. It's narrower in what it covers, Google Search's AI features specifically, not ChatGPT or Claude, but for the surfaces it does cover, it's ground truth rather than inference.
The other update: vetting who you hire
On June 5, 2026, Google updated its long-standing "Do you need an SEO?" hiring guide, the page it has maintained for years to help site owners evaluate SEO vendors, to add a section specifically about AEO and GEO as a service category. The update tells buyers to check whether a vendor's AEO/GEO recommendations actually align with Google's own published guidance, and separately warns that no third-party tool has access to Google's internal ranking data and none can guarantee performance. It also cautions against vendors who imply they're somehow "approved" by Google.
That warning exists because it needed to. The past eighteen months produced a wave of vendors selling llms.txt implementation, AI-specific content rewrites, and citation-tracking dashboards as though they were the entire discipline, sometimes with pricing that implied a level of certainty about AI-search ranking mechanics nobody outside Google actually has. Google's own documentation now gives buyers a specific, citable reason to ask a harder question in the next vendor pitch: does this match what Google itself says is necessary?
The "no third-party tool has access to Google's internal ranking data" line is worth sitting with a bit longer than a warning label deserves. It's true of every AI-visibility platform on the market right now, not a knock on any one vendor specifically. Every citation-tracking tool, every mention-rate dashboard, everything built to estimate how often ChatGPT or AI Overviews cite a given brand, is inferring from sampled prompts and observed outputs, not reading a Google system directly. That's not disqualifying; it's the only method available to anyone outside Google or OpenAI. But it does mean a vendor pitching "guaranteed AI visibility" is making a claim the underlying tooling can't actually back, and Google just put that in writing.
This is also where the two documents reinforce each other rather than standing alone. The mythbusting guide tells you which tactics don't need to be on an invoice. The hiring-guide update tells you how to evaluate the person or firm sending that invoice. A vendor who leads a pitch with "we'll build your llms.txt and optimize your content structure for AI crawlers" is proposing exactly the two line items Google's guide says aren't required, priced as though they're the whole engagement. That's now a two-minute check against Google's own public documentation, not a judgment call you have to make on trust alone.
Where this leaves the GEO industry
None of this means generative engine optimization is fake or that AI search visibility doesn't matter. It means the tactics that actually move it were mostly always SEO tactics: extractable structure, demonstrated authority, and content worth citing, the same three signals we described in the anatomy of an AI citation back in July. What Google's guide does is take the file-format cosplay out of the conversation and leave the actual work standing. Schema that AI engines read still matters, not because it's a special GEO tactic but because Organization, Article, and author schema were already useful structured-data practice before AI Overviews existed.
That's a useful filter for evaluating your own generative engine optimization plan, or anyone else's. If a recommendation only makes sense in a GEO-specific frame, chasing a file format or a rewrite that has no other justification, Google's own documentation now gives you grounds to question it. If it's a recommendation that would already improve your core SEO foundation, crawlability, content quality, structured data you can defend, it survives the mythbusting section intact, because that's the work Google says was never optional in the first place.
Do this next
Pull whatever GEO-specific line items are sitting in your current budget, an llms.txt retainer, a content-chunking rewrite, a schema package sold as "AI-specific," and check each one against Google's own five-item mythbusting list above. Anything that only exists to satisfy an AI crawler and has no independent SEO justification is a candidate to cut or renegotiate. What survives that filter, technical crawlability, content that states a real position, structured data grounded in actual SEO best practice, is the work worth funding, whether you're serving B2B SaaS buyers or any other category where the buying committee now researches through both a search bar and a chat window.
None of this makes generative engine optimization a solved problem or a discipline that doesn't deserve dedicated attention. It makes the honest version of that work easier to defend, because the fundamentals it actually depends on now have Google's name attached in writing, and the tactics that were always closer to theater have Google's name attached to the debunking.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Tyler leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.