Google published its first official guide to generative AI optimization this week, and the most important line in it is the one that sounds least exciting: AEO and GEO are still SEO.
No special schema. No llms.txt file. No content chunking strategy. Google's own guidance says those tactics are unnecessary for most publishers. The agencies who sold you a "GEO audit" built on those three things owe you a refund.
But read carefully. "Still SEO" doesn't mean nothing changed. It means the foundation matters more than ever, and most service businesses are still building on sand. Here's what the guide actually says, what it leaves out, and what a $5M home services company or 12-location dental group should do about it.
What Google actually said
The Google Search Central blog post, published May 15, is short for something this consequential. The core message: the same signals that earn classic organic rankings also earn citations in AI Overviews, AI Mode, and other generative features. E-E-A-T is the engine. Content quality is the fuel. Structured data helps when it's relevant, but there's no magic AI-specific markup that unlocks citation.
Google specifically called out three tactics circulating in the SEO community and dismissed them. First, llms.txt. The robots.txt cousin that some SEOs started recommending as a way to "talk to AI crawlers" directly. Google says it doesn't use it for Search. Second, content chunking. The idea that you should restructure pages into AI-digestible blocks. Google says write for humans. Third, special schema for AI. There is no schema type that makes Google's AI more likely to cite you. The existing structured data vocabulary still applies where it's always applied.
That's clarifying. What the guide doesn't say is equally important. It doesn't say that optimizing for ChatGPT, Perplexity, and Gemini follows the same rules. Google can speak for Google. The other AI engines have different training pipelines, different citation logic, and different relationships with structured data. A complete GEO strategy covers all five surfaces, not just AI Overviews.
Why E-E-A-T is the whole game
Google's framework for content quality has four components: Experience, Expertise, Authoritativeness, Trustworthiness. Most service businesses treat E-E-A-T like a checklist item. A few author bios. An "About" page with credentials. A handful of reviews. That approach gets you to average. Average doesn't get cited.
AI engines behave like a senior editor at a trade publication deciding who to quote. They pull from sources with documented track records, consistent topical depth, and third-party corroboration. A single great page doesn't move the needle. A site where every page on HVAC maintenance reflects the same expert voice, links to primary sources, and gets mentioned in local press and industry directories, that site gets quoted.
Brand mentions are a signal that most service businesses underestimate. When a local news outlet covers your company, when a trade association lists you as a resource, when a satisfied customer posts a detailed review on a platform other than Google, those mentions become part of the corroboration web AI engines use. GEO is to SEO what Spotify was to record stores. Same job of getting your content found. Structurally different reward system underneath.
“AI engines behave like a senior editor deciding who to quote. One great page doesn't move the needle. A documented track record does.”
Structured data still matters
GA4 is now tracking AI assistant traffic as a distinct channel. That landed this week too, buried under the louder news. For the first time, you can segment traffic arriving from ChatGPT, Perplexity, and Gemini referrals separately from organic search. If you haven't set up that reporting yet, you're flying blind on roughly 15 to 20 percent of your inbound discovery.
On the same day, Google confirmed that FAQ rich results are gone from Search. Not from structured data. From the visible SERP feature. The underlying FAQ schema you've already deployed still gets processed, still feeds AI answer extraction, and still helps Google understand the question-and-answer relationships on your pages. Removing FAQ schema in response to this news is the wrong call. Keep the markup. Lose the expectation of the accordion on the SERP.
The structured data types that matter for AI citation are the ones that establish identity and context, not the ones that trigger visual features. Schema for your organization, your services, your geographic coverage, your team members' credentials, your reviews. These are the signals that tell AI engines who you are, what you do, and why you're credible. Think of it less like decorating a page and more like filling out a professional profile that every AI in the world can read.
The schema types worth maintaining for AI visibility
- Organization schema with full NAP consistency, logo, and sameAs links to your social and directory profiles
- LocalBusiness schema on every location page, including service area, hours, and review aggregate
- Service schema that names specific services with descriptions, not just a generic category
- Person schema for team members and authors tied to the content they produce
- FAQ schema on pages where genuine question-answer content exists, even though the rich result is gone
- Review and AggregateRating schema sourced from verified third-party platforms
The local visibility problem
If you run a multi-location service business, the AI citation problem has a geographic layer that most SEO guides skip. AI engines are terrible at citing thin location pages. The 47-word city landing page that was a passable SEO play in 2021 is invisible to Perplexity in 2026.
The service businesses getting cited in local AI answers share a specific profile. Their location pages have genuine depth: real team bios for that location, actual customer stories from that market, service descriptions that reference local context. Their Google Business Profiles are complete and active. Their review velocity is consistent, not a burst of 40 reviews in January followed by silence. And critically, they appear on local news sites, community directories, and industry associations as genuine entities, not just as businesses that paid for a listing.
That last part is the hard work. It's also the moat. Bolt-on AI tactics are the screen on a 2019 Honda dashboard. Real local authority is the Tesla. One looks like the future. One actually runs differently underneath. You can read more about the mechanics of local AI visibility in our SEO & GEO service overview.
The five surfaces, not one
Here's where Google's guide is useful but incomplete. It describes how to earn visibility in Google's AI features. That's one surface. A business getting discovered through AI in 2026 needs to appear across five distinct surfaces. Google AI Overviews. ChatGPT search. Perplexity. Gemini. Claude via the browser and operator integrations.
Each surface has different training data, different update cycles, and different citation patterns. Perplexity tends to favor pages with dense, verifiable facts and clear sourcing. ChatGPT's web search favors recency and structured formatting. Gemini, being Google-owned, leans hardest on the same E-E-A-T signals. Claude's operator integrations pull from whatever context the operator provides, which is a different problem entirely. Our GEO strategy framework breaks down the citation logic for each engine.
- 01Audit your E-E-A-T foundation first. Author bios, credential documentation, third-party mentions. This is the prerequisite for everything else.
- 02Set up GA4 AI assistant traffic segmentation this week. You can't optimize what you can't measure.
- 03Audit your structured data against the six schema types listed above. Not for rich results. For AI identity signals.
- 04Rebuild location pages that are under 400 words of genuine, local-specific content. Thin pages are invisible to AI.
- 05Start a brand mention acquisition program. Local press, industry directories, podcast appearances. The corroboration web is what gets you cited.
Google's guide is a gift in the way that a clear diagnosis is a gift. You might not love the news, but at least you know what you're treating. The businesses that win AI citations over the next 24 months aren't the ones who find a technical shortcut. They're the ones who build genuine authority that no algorithm update can ignore.
That's a longer play than most agencies want to sell. It requires consistent content, active review management, real PR, and structured data maintained as a living system, not a one-time setup. It's closer to building a Patagonia brand reputation than running a Facebook ad campaign. Slower, stickier, and much harder to copy.
The businesses asking "how do we rank in AI search" are asking the right question six months late. The better question is: when a user asks an AI engine who the best HVAC company in Phoenix is, or which accounting firm handles multi-state construction payroll, is there enough corroborated evidence on the web that the AI would confidently cite you? If the honest answer is no, that's the gap to close. We're building that case for clients right now through our SEO & GEO work, and the window to build the moat before competitors wake up is genuinely narrow.