Why SEO isn't obsolete: AI search still runs on the ranking system

The logic is simple: these generative AI features are rooted in Google's existing core search ranking and quality systems. AI stands on the foundation SEO already built, picking and using content through two mechanisms—it didn't start from scratch with some new way of understanding web pages.

RAG: the AI looks things up before it answers

RAG stands for retrieval-augmented generation; Google also calls it grounding. Here's what it means: before answering you, the AI leans on the core ranking system to retrieve relevant, sufficiently fresh pages from Google's index, reviews the specifics on those pages, then generates a more reliable answer—complete with clickable, prominent links back to those pages.

That chain has a direct implication: whatever AI cites in its answers comes from pages that were crawled and indexed. If your page never makes it into the index, AI can't use it in an answer either. So the old SEO homework—being crawlable, being indexable—hasn't been rendered obsolete by AI search at all.

Query fan-out: one question becomes several searches

The second mechanism is called query fan-out. Ask one question, and the AI automatically generates a set of related sub-questions, fires them off as searches simultaneously, pulls back more relevant results, and pieces the information together to answer you.

For content creators, this means: even if a user's original phrasing doesn't match the wording on your page, your content can still get picked up by one of the fanned-out queries, as long as it genuinely answers a related sub-question. There's a clear line not to cross here, though: don't respond to this by building a separate page for every conceivable phrasing and padding your page count to game visibility. That violates the "scalable content abuse" spam policy—more pages don't make a site more useful or more relevant, and it won't pay off long-term.

As for all the AEO and GEO talk floating around, Google's position is clear: from Search's perspective, optimizing for generative AI search just means optimizing the search experience, which is still SEO. If you're thinking about buying third-party AEO/GEO advice or services, check it against Google's own guidance on evaluating third-party SEO advice before you pay for anything.

5 things you don't need to do: "AI optimization" buzzwords, named and dismissed

This is the most reassuring part of the whole guide. Google lines up the most widely circulated "AI optimization" tactics and tells you, one by one, that as far as Google Search is concerned, you can safely ignore them.

So what should you actually do: write what nobody else can

Having crossed out the buzzwords, Google also spells out where the effort should actually go. The item at the top, with the biggest impact of all: create unique, valuable content that can't be found anywhere else.

Google uses a strikingly concrete comparison to illustrate what "worth writing" actually looks like.

What does "good content" actually look like? Google boils it down to four traits—also the traits that neither other people nor AI can easily copy.

Yes, it's okay to use AI to help write content

This is the question most people agonize over, and Google gives a clear answer: yes, you can use generative AI tools to help create content, as long as the finished product meets Search Essentials and the spam policies. The tool itself isn't the problem—low-quality, cobbled-together content made purely to game rankings is.

The technical foundation still matters: AI needs to read your site first

No matter how good your content is, AI still has to find and understand your page first. The first step in the RAG chain is "retrieve from the index," so whether you can be crawled and indexed is the precondition for AI ever using you. None of the old SEO homework is obsolete—Google lists six items.

Local businesses and e-commerce: don't forget your merchant data

Generative AI answers will surface product and local business information where relevant. If you run a local business or an e-commerce operation, filling out your product and store details through Merchant Center (like a product data feed) and Google Business Profile lets that information show up in both AI answers and regular search results. There's also Business Agent, a conversational experience that lets customers chat directly with your brand right inside Google Search.

AI agents are starting to browse your site on their own

The guide also flags an emerging trend: AI agents—systems that can autonomously complete tasks on a person's behalf, like booking a table or comparing specs across products.

These agents come in different forms; one type is the browser agent, which visits your site directly and gathers what it needs in a few ways: analyzing the page's visual rendering (like a screenshot), inspecting the DOM structure (the page's underlying skeleton code), or reading the accessibility tree (the structural information screen readers and similar tools rely on). Going forward, the entities "reading" your site won't just be people and search crawlers—AI acting on people's behalf will be in the mix too.

If this is relevant to your business and you have the bandwidth, it's worth checking out Google's list of agentic experiences and its guide on agent-friendly website best practices, to get a rough sense of how your site should prepare for today's browser agents. One direction now emerging: protocols like UCP (Universal Commerce Protocol) are starting to appear, which will let agents in search do more.

How to measure results—and don't trust tools claiming "insider metrics"

After doing all this, how do you know it's working? There's really only one legitimate path: the Generative AI performance report in Search Console, which shows how your content performs in Google Search and Discover's AI features, and how people are discovering you through them.

Here's a bucket of cold water, too: be wary of any third-party tool promising ranking success or claiming to use Google's "internal" metrics. No third-party tool has access to Google's internal ranking or AI systems. Use these tools if they genuinely fit your workflow, but check their advice against Google's official guidance.

plenty of content thrives in Google Search (including generative AI experiences) without any overt SEO at all. Google Search Central, original text

The guide ends on a modest note: you don't need to check off every single item here to succeed. Plenty of content does fine in search (AI experiences included) without any conspicuous SEO at all. Make content for people, and the rest will mostly follow.

Google Search's own team, Search Central, published a guide that says it plainly: SEO isn't obsolete in the age of AI search, and a lot of the new "AI optimization" tactics are a waste of effort.

So how does AI actually use your web page? Two mechanisms explain it—and explain why SEO isn't obsolete along the way.

These AI features don't conjure answers out of thin air. RAG (retrieval-augmented generation): the AI first searches Google's index for relevant pages, generates an answer based on them, and attaches clickable source links. Query fan-out: ask one question, and it automatically splits that into several related sub-questions to search at once, then pieces together an answer.

With the buzzwords crossed out, Google names both where to spend effort and where not to. In one line: write what neither other people nor AI can write.

Google's sharpest comparison: "7 Tips for First-Time Homebuyers"—anyone could write it, AI can generate it ✘; "Why We Skipped the Home Inspection and Saved Money: The Truth About Our Plumbing"—first-hand experience, impossible for anyone else to copy ✓. (Non-commoditized content = not the kind of generic stuff that shows up in every search.)

(One emerging variable: AI agents are starting to browse sites on people's behalf, reading screenshots and page structure to gather data. Worth keeping an eye on for e-commerce and local businesses, though it's still early days.)