Google Just Told You to Stop Doing AEO. Here's What That Actually Means.

On May 15, Google updated its Search Central documentation with an official guide to optimizing for AI Search. The headline is blunt: AEO and GEO are just SEO.

That's not a dismissal. It's a clarification that SEO managers, marketing directors, and anyone currently navigating the AEO waters needed to hear. Here's a breakdown of what Google said, what it debunked, and where the genuinely new ground is.

The Mechanics Behind AI Overviews

Google's generative AI features run on the same ranking and indexing infrastructure as regular Search. Two techniques drive this:

RAG (Retrieval-Augmented Generation): The AI pulls content from Google's index to ground its answers, then surfaces clickable source links alongside the generated response.

Query fan-out: When you ask a question, the model generates multiple related queries in parallel before composing an answer. It's not pulling from one source: it's synthesizing across several.

This matters because it means your content's path into AI answers runs through the same crawling, indexing, and ranking signals it always has.

What Google Says You Should Do

The advice here isn't particularly surprising, but the framing is useful.

Create content that couldn't exist anywhere else. Google explicitly calls out "commodity content", which is the kind of piece that regurgitates what's already been written. Their example: "7 Tips for First-Time Homebuyers" is common knowledge with no unique angle. A piece titled "Why We Waived the Inspection and Saved Money: A Look Inside the Sewer Line" is a first-hand account that adds something to the record. The AI has less reason to pull from you if you're saying the same thing as fifty other sites.

Keep your technical fundamentals clean. Indexable pages, crawlable content, a snippet that can be extracted, minimal duplicate content. The basics still apply.

Use your existing merchant signals. For e-commerce and local, Merchant Center feeds and Google Business Profiles feed directly into AI answers. Nothing new here, but worth auditing if you haven't.

The Debunks Are the Most Useful Part

Google is direct about a set of tactics that have picked up real traction in the AEO/GEO space:

llms.txt files: Not needed. Google treats them like any other indexable file, which means no special weighting, no preferential treatment.

"Chunking" content: Not needed. Google's systems already handle pages that cover multiple topics. You don't need to restructure your content architecture to match how you imagine the AI parses it.

Rewriting content for AI queries: Not needed. The models understand synonyms and intent. Chasing every long-tail variant of a query is wasted effort.

Chasing inauthentic mentions: Ineffective. Google's spam systems filter these out. Tactics built around manufacturing brand mentions without substance don't move the needle.

Adding structured data specifically for AI Search: Structured data isn't required for AI answers. Google's systems don't need it to understand your content for that purpose. It still earns rich results in traditional Search, so keep implementing it for those reasons. Just don't expect it to move your AI visibility numbers.

A category of services emerged over the past two years selling AI-specific optimization work: llms.txt configuration, content chunking audits, AEO-specific structured data overlays, AI mention-building campaigns. These services were built on the premise that AI Search operates on different rules than regular Search and therefore requires different inputs. Google's guide is a formal statement that this premise is wrong. The AI features inherit the same crawling, indexing, ranking, and spam systems. There is no separate track.

The Part Worth Paying Attention To

The guide's section on agentic experiences is where things get genuinely forward-looking, and it's structurally different from everything else in the document. The rest is clarification of existing systems. This section describes something that doesn't fully exist yet but is being built.

AI agents can now access websites autonomously to complete tasks. Where current AI Search retrieves and summarizes, agents act: they can book, compare, purchase, fill forms. They do this by analyzing the visual rendering of a page, its DOM structure, and the accessibility tree simultaneously. They're not reading your content the way a crawler does. They're parsing your interface the way a user does, then deciding what's actionable.

This matters because it shifts the relevant question. For AI Search, the question is whether your content gets cited. For agentic AI, the question is whether your site is legible enough for an agent to complete a task on it without failing. A poorly labeled button, an inaccessible modal, a checkout flow that breaks under programmatic interaction — these become the equivalent of a 404 for an agent trying to act on your behalf.

Google references the Universal Commerce Protocol (UCP) in this context. UCP is an emerging standard designed to let Search agents not just retrieve information but take actions: placing orders, making reservations, completing transactions. It's early, and adoption is limited. But the direction is clear: Google is building toward a web where the agent completes the task rather than handing the user off to complete it themselves.

The shift from retrieval to action changes what "visibility" means. Getting cited in an AI answer is a passive win. Getting selected by an agent to complete a transaction is something else. Which sites agents trust, which interfaces they can navigate, which commerce flows they can complete without breaking, that's the competitive surface that UCP is designed to define. It's worth watching how quickly it moves from reference to requirement.

The Honest Reading

Google's message is: stop looking for AI-specific shortcuts and do the SEO work that produces results everywhere.

For practitioners already operating at a high level, this validates what they've been doing. For teams that have been chasing AEO-specific tactics, it's a signal to audit where time is actually going.

Validation Doesn't Equal Visibility

The inputs Google describes are the same ones SEO teams have always optimized for. What's changed is the output layer. Traditional rank tracking shows you position in a list of blue links; it doesn't show you whether you're being cited in an AI Overview, surfaced in AI Mode, or pulled into a fan-out response for a related query. The work hasn't changed. The way that work shows up has.

That's the gap tools like Operyn close: tracking which queries mention you, which pages get cited, and how your share of voice in AI-mediated results compares to competitors. Google can tell you what inputs matter. You still need a way to see how those inputs are landing in the surface that's increasingly mediating your customers' attention.