Google Just Confirmed SEO Still Matters in the AI Era. Here's What They Actually Said.

For the last two years, every conference I've sat through has had at least one slide claiming SEO was over. The reasoning is always the same. AI Overviews are eating the click. ChatGPT is the new front page. We need a whole new playbook: AEO, GEO, llms.txt, "content for the LLMs."

I run a marketing agency that builds and maintains websites for small businesses. We don't get to chase frameworks because they're trending. If we change how we work, the change needs to be grounded in something more durable than a LinkedIn carousel. So when Google published its own AI optimization guide for search this month, we read it ourselves rather than waiting for the takes to come in.

The short version: Google directly addressed whether SEO is still relevant, and the answer is yes. Quoting the guide:

> "In short, yes! The best practices for SEO continue to be relevant because our generative AI features on Google Search are rooted in our core Search ranking and quality systems."

Read that line carefully, because almost everything else in the document follows from it. AI Overviews are not a different system competing with classic search. They are built on top of it. Same crawl, same index, same ranking signals, with a generative layer on the front.

That changes the conversation about what to do next, and it especially changes the conversation small business owners are being sold right now. Let's go through what's actually new, what stays the same, and what to do about it.

What's actually new

Two mechanisms are worth understanding because they explain how an AI surface selects content.

The first is retrieval-augmented generation, or RAG. Google describes it as "a technique...used to improve the quality, accuracy, and freshness of AI responses by relying on our core Search ranking systems to retrieve relevant, up-to-date web pages." Plain language: when an AI Overview answers a question, it isn't pulling from a static training set. It's running a retrieval pass against the same web index that powers regular Search and then composing an answer from the pages it found.

The second is query fan-out: "A set of concurrent, related queries generated by the model to request more information and fetch additional relevant search results." When a user asks a complex question, the model breaks it into several smaller queries, hits the index for each, and stitches the results back into a single response.

What this means in practice: a single page rarely "wins" an AI Overview the way a single page can rank #1 for a keyword. Multiple pages contribute. The pages that get pulled into more of those fan-out queries get cited more often. The work isn't to write one perfect AI-optimized article. The work is to be the trusted source on enough adjacent questions in your space that you keep getting retrieved.

What's stayed the same (and what Google explicitly told people to stop doing)

The guide is unusually direct about the new vocabulary the AI optimization industry has built around itself. Google calls it out by name. The guide warns against "AEO/GEO hacks" — things like llms.txt files, content chunking schemes, and artificial mentions designed to game LLM ingestion. The position is clear: these don't enhance visibility.

I'd add a builder's note here. None of those tactics ever made sense in the first place. The big LLMs do not respect llms.txt as an ingestion contract. Content chunking only "helps" if you're optimizing for a hosted RAG system you control, which is not what consumer AI search is. The hacks were a vocabulary problem disguised as a strategy. Google saying so out loud just shortens the cycle.

On the substance, the fundamentals are unchanged:

• Technical crawlability and indexation requirements. A site that Google can't crawl can't appear in the index, and a page not in the index can't be retrieved by the RAG layer.
• Content quality standards: helpful, reliable, people-first content. This is the same E-E-A-T conversation we've been having for years.
• Duplicate content and spam policies. AI generation of pages-at-scale to chase keyword variants is still spam by Google's definition, and the guide reaffirms it.
• Page experience and Core Web Vitals. Slow, broken, mobile-hostile pages are still penalized.

Google's framing of the bottom line: "All existing technical SEO best practices continue to be worthwhile."

Two specifics worth flagging

There are two details in the document that I think are getting under-discussed.

Structured data isn't a special AI weapon. A lot of consultants have been selling schema markup as the magic wand for AI Overviews. The guide is explicit: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." Schema still matters — it makes pages eligible for rich results, it helps disambiguate entities, it's still part of doing things properly. But it isn't the unlock people are being sold.

Unique, first-hand content is favored over recycled summaries. The guide pushes hard on "unique, non-commodity content" with expert perspective over generic synthesized information. If your content strategy was already "rewrite the top three results and add an FAQ block," that approach has a shorter shelf life now than it did a year ago. RAG is good at finding ten lightly-paraphrased versions of the same article. It will tend to surface the original one, or the one that adds something the others don't.

What this means if you run a small business website

If you own a 10–50 page small business website and you've been hearing that you need to rebuild the whole thing for AI, here is the honest read.

You don't need a new framework. You probably need a technically sound site, content that actually answers the questions your customers are asking, and enough internal structure that Google's crawler can find everything in a reasonable budget. That is what we'd have told you to do five years ago, and that is what the guide tells you to do now.

The specific technical work that still pays: verifying the site in Search Console so you can see what's getting indexed, following JavaScript SEO best practices if you're on a framework like Next.js or React, optimizing crawl budget for sites that update frequently, using semantic HTML for accessibility and clean parsing, and keeping page latency and mobile experience tight. None of this is new advice. All of it was already correct.

The new work, if there is any: auditing whether your category's queries are getting AI Overviews, and if so, whose content is getting cited. If your competitors are getting pulled into the AI summary and you aren't, the gap is almost always upstream — content depth, originality, indexation coverage, or a Search Console-level issue you haven't looked at recently. It's not a missing llms.txt file.

The work to stop paying for: anyone selling you "AI optimization" as a category-distinct service with new acronyms attached. The vocabulary is new. The work underneath is the same SEO foundation that's been correct the entire time. If the proposal you're reading is mostly new jargon and a higher invoice, that's the trade you're being asked to make.

A note on the broader pattern

This is the third time in the last two years I've watched the marketing industry generate a new layer of consulting around a real but smaller change in how search works. Featured snippets did it. Voice search did it. AI Overviews are doing it now. The pattern is consistent: a real shift happens, the existing technical fundamentals continue to be most of what matters, and a service layer grows up around the shift that overstates how much of what came before is now broken.

Google publishing this guide is useful because it lets owners and operators check the influencer takes against the primary source. If you're being told that AI search has rewritten the rulebook, the rulebook itself just confirmed the opposite.

The work for a small business website in 2026 looks a lot like the work for a small business website in 2023: a fast, accessible, well-structured site; content that actually answers the question; a real point of view that doesn't read like every other page on the topic; and a team that reads the primary source when something changes.

That's the kind of source-grounded technical work we do for clients at Commonwealth Creative. If you want a second pair of eyes on your site, we're happy to take a look.