Google published an official guide this week called Optimizing your website for generative AI features on Google Search. It covers AI Overviews, AI Mode, and what Google actually wants from sites that want to show up in AI-generated responses. It's worth reading in full. It's also worth having someone cut through the parts that are polite corporate hedging versus the parts that actually change what you should be doing. That's what this post is for.
The short version: most of what Google recommends is foundational SEO you should already be doing, with one genuinely useful section on what not to bother with. There are also a few places where the guide quietly says something important between the lines.
SEO Still Matters, and Here's Why That's Not Obvious
The first thing Google addresses is whether traditional SEO is still relevant now that AI-generated answers are becoming the default response to a huge portion of queries. Their answer is yes, and the reason is technical: AI Overviews and AI Mode use retrieval-augmented generation (RAG), which means they're pulling from the Search index, not generating answers from scratch. If a page isn't indexed and ranking, it isn't being considered as a source for AI responses.
This is important because a lot of the "AI search is different, throw out your SEO playbook" content circulating right now is wrong, or at least premature. Google's AI systems are downstream of Google Search ranking systems. Getting your content into AI responses still starts with getting it indexed and evaluated well by the core ranking systems. The AI layer is built on top of the same infrastructure, not parallel to it.
There's also a concept called query fan-out that's worth understanding. When someone asks a complex question, the model generates a set of related sub-queries and pulls results for each. If your content answers a narrow version of a broader question, you can get surfaced in AI responses for queries that don't exactly match your page. That's a meaningful shift from traditional keyword matching, and it rewards content that covers a topic with depth rather than content that's optimized for a single phrase.
What Google Actually Wants from Your Content
The guide spends the most time on content quality, and the framing is clearer than most SEO guidance you'll read. Google distinguishes between commodity content and non-commodity content, and the distinction is useful. Commodity content is stuff anyone could write: "7 Tips for First-Time Homebuyers." Non-commodity content is stuff that comes from actual expertise or experience: "Why We Waived the Inspection and Saved Money: A Look Inside the Sewer Line."
The practical implication for brand content is significant. Product descriptions, category pages, and blog posts that simply restate common knowledge are not going to perform well in AI-mediated search, regardless of how well they're technically optimized. AI systems are looking for unique perspectives, specific expertise, and content that goes beyond what could be produced by a generative AI model from scratch. If your content reads like it was written by AI summarizing other content, it won't be the source AI chooses to cite.
This is the core argument for investing in content optimization for AI search: the work isn't just technical. It's about building content that carries the kind of authority and specificity that AI systems treat as worth quoting. Generic content, however well-structured, doesn't make that cut.
The Mythbusting Section Is the Most Useful Part
Google dedicates a full section to things you don't need to do, and it's refreshingly direct. A few things they explicitly say to ignore: llms.txt files and other special AI markup files, "chunking" content into tiny pieces for AI consumption, rewriting content specifically for AI systems using different language patterns, seeking inauthentic brand mentions across the web, and overfocusing on schema markup specifically for AI visibility.
This matters because there's a growing industry of consultants and tools selling all of these things. The llms.txt file in particular has gotten a lot of traction, with various vendors claiming it helps AI systems understand your site. Google says flatly that they don't treat it in any special way. If you've been told to add one, it's probably not hurting you, but it's also not doing anything.
The structured data point is nuanced: Google says it's still worth using as part of an overall SEO strategy because it helps with rich results, but there's no special schema markup that boosts AI visibility. If someone is pitching you a new schema type specifically designed to improve your AI Overview performance, that's not a thing.
Technical SEO Still Has to Be Right
Google is clear that the technical requirements for appearing in AI features are the same as for appearing in standard Search: the page has to be indexed, eligible to show a snippet, and meet core technical requirements. Pages blocked from crawling, pages with noindex directives, and pages that can't be rendered are simply not in the pool for AI responses.
The guide also flags JavaScript SEO as a complexity worth addressing. If your site relies heavily on client-side rendering and your content only becomes visible after JavaScript executes, you may have indexing gaps you're not aware of. Google can process JavaScript, but it's more resource-intensive and less reliable than static HTML. For e-commerce brands with product and category pages that are heavily JS-rendered, this is worth auditing.
Page experience also gets a mention: fast loading, good mobile rendering, and easy-to-identify main content. None of this is new, but the guide reinforces that it's still part of the equation. A technically clean site isn't sufficient on its own, but a technically broken site can prevent otherwise strong content from getting surfaced at all.
The Agentic Section Is the One to Watch
The most forward-looking part of the guide covers agentic experiences, and it's short because this is still early. AI agents that can take actions on behalf of users, booking, comparing, purchasing, are increasingly interacting with websites directly, not just reading indexed content. Google points to the Universal Commerce Protocol (UCP) as an emerging standard for how agents will interact with sites, and recommends reviewing their agent-friendly website best practices guide.
This is further out than AI Overviews, and Google is appropriately cautious about it. But for brands in e-commerce, it's worth knowing the direction of travel. The sites that are structured well for AI agents, clear product data, accessible cart and purchase flows, consistent entity information, will have an advantage when agentic shopping becomes mainstream. Building for that future doesn't require doing anything different today beyond getting the fundamentals right.
If you want a clear read on where your site stands against the standards Google laid out in this guide, and what the fastest path to improvement looks like for your specific situation, schedule a call with us. We work through exactly these questions with brands every week and can tell you what's worth prioritizing and what you can safely ignore.