AI Overviews and AI Mode: what founders should change

The question founders keep asking right now is some version of: "What do I need to change for AI Overviews?" Often right after someone offered them a "GEO audit."

The honest answer, from Google's own documentation, is: less than you've been told. But not nothing. There are a few real changes, a new control you should know about, and a new report that finally lets you measure this.

Everything below comes from Google's docs, with links, so you can check it yourself.

Short answer: Google says AI Overviews and AI Mode run on its core Search ranking systems, so normal SEO still applies. Pages need to be indexed and snippet-eligible. You don't need llms.txt, special schema or "chunked" content. What helps is unique, non-commodity content. Measure it with Search Console's Generative AI performance report.

How AI Overviews and AI Mode pick links

Google's AI optimization guide, last updated in July 2026, describes two techniques.

Retrieval-augmented generation, or grounding. Google's AI features use "our core Search ranking systems to retrieve relevant, up-to-date web pages from our Search index." The model then writes a response and shows "prominent, clickable links to relevant web pages that support the information."

Query fan-out. The model generates "a set of concurrent, related queries" to fetch more results. Google's example: for "how to fix a lawn that's full of weeds," fan-out queries might include "best herbicides for lawns" and "remove weeds without chemicals." Google's AI features page says this lets the features show "a wider and more diverse set of helpful links" than classic results.

Two things follow for founders:

1. If you don't rank in regular Search, you're unlikely to be grounded on. The retrieval step uses the same index and ranking systems.
2. You can be cited for a query you don't rank for directly. Fan-out means a page that answers a sub-question well can be pulled into an answer for a broader query.

Our glossary has short entries on grounding, query fan-out and AI Overviews.

The eligibility checklist

Google's requirements are short. To appear as a supporting link in AI Overviews or AI Mode, a page must be:

• Indexed and eligible to be shown in Google Search with a snippet. That's the same bar as a normal result with a description.
• On a site included in Search generative AI features in Search Console. This is the default, and it's a new setting we cover below.

That's it. Google's AI features page says: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary."

So the first real change is a check, not a project: make sure your important pages are indexed and nothing blocks snippets. A stray nosnippet, a max-snippet:0, or a page Google won't index takes you out of AI features along with everything else. Run key pages through our meta tag checker and look at the Page indexing report. Our Search Console guide shows where.

What Google says you can ignore

The guide has a section literally called "Mythbusting generative AI search." For Google Search, you can skip:

• llms.txt and "special" files. "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search," because "Google Search itself doesn't use them." Keeping one "will neither harm nor help" your visibility. (We wrote about llms.txt in more detail. It can still be useful for other agents.)
• Chunking. "There's no requirement to break your content into tiny pieces for AI to better understand it."
• Rewriting for AI. "You don't need to write in a specific way just for generative AI search. AI systems can understand synonyms and general meanings."
• Special structured data. "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." It's still worth using for rich results.
• Inauthentic mentions. "Seeking inauthentic 'mentions' across the web isn't as helpful as it might seem," because the AI features depend on the same systems that "block spam."

And on terminology: "From Google Search's perspective, optimizing for generative AI search is optimizing for the search experience, and thus still SEO."

The one thing that could hurt you

The guide warns against a tactic that's being sold hard right now: making a page for every variation of a query, including fan-out queries.

In Google's words, creating separate content "for every possible variation of how people might search (for example, by focusing on other queries that people have asked, or fan-out queries)" primarily to manipulate rankings or AI responses "violates Google's scaled content abuse spam policy." It adds that "a high quantity of pages doesn't make a website higher quality."

If a tool or agency's AI plan is "generate 500 pages targeting fan-out queries," that's the plan Google describes as spam. We covered how to build page families that stay indexed in programmatic SEO without getting deindexed.

What actually helps: non-commodity content

The guide says content that's "unique, compelling, and useful will likely influence your website's presence in generative AI search in the long run more than any of the other suggestions."

Its example is useful. Commodity content is something like "7 Tips for First-Time Homebuyers," which is "often based on common knowledge." Non-commodity content, like "Why We Waived the Inspection & Saved Money: A Look Inside the Sewer Line," gives "unique expert or experienced takes that go beyond common knowledge."

For a startup, that points to content you're uniquely able to write:

• What you learned building the product. Our story of getting throttled after 2,500 posts is that kind of post.
• Honest comparisons with the products people evaluate you against, with facts you've checked.
• Real data from your product or your market that nobody else has.
• Specific how-tos for problems your users actually hit, with screenshots of your product doing it.

A model can generate the generic version of any article. It can't generate your experience. That's the gap Google says it wants to fill.

The new control: Search generative AI

In June 2026, Google added a setting to Search Console. Its help page describes three options under Settings → Search generative AI:

• Include (the default): your content can appear in, and help ground, AI Overviews, AI Mode and generative AI features in Discover.
• Exclude: your content won't be shown or linked in those features, and "you won't receive any traffic or impressions from these features."
• Inherit: a child property follows its parent's setting.

Google says the control "isn't used as a ranking or inclusion signal affecting other parts of Search." Changes take effect within one to two days, and some content can take longer because of caching.

Should a startup exclude itself? Almost never. For a new product, being linked in an AI answer is visibility you can't easily get elsewhere. The control is mainly for publishers who've decided the trade isn't worth it for them. But check the setting is on Include, especially if someone else set up your Search Console.

Also note what the control is not. It's separate from Google-Extended, the robots.txt token for Gemini training and grounding in other Google products. Google's crawler docs say Google-Extended "does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search." Our guide to robots.txt, llms.txt and AI crawlers covers the rest.

How to measure it now

Two places, both free:

The Performance report. Google's AI features page says traffic from AI Overviews and AI Mode is "included in the overall search traffic in Search Console" under the Web search type. It isn't separated out.

The Generative AI performance report. Rolled out to all sites on August 31, 2026, per Google's help page. It shows impressions from AI Overviews and AI Mode by page, country, device and date. Two links to your site in one AI response count as a single impression. There are no clicks and no queries in it yet.

What we'd do with it:

1. Note which pages get AI impressions. Those are the pages Google's AI already trusts on your site.
2. Compare them with your top pages in the regular Performance report. Pages strong in one and weak in the other are worth a look.
3. Improve the pages that get AI impressions but few clicks: sharper titles, clearer first paragraphs, a reason to click through for detail the summary can't give.

For ChatGPT, Perplexity and others, there's no equivalent report. We explain a manual method in how to measure AI visibility without paid tools.

What founders should actually change

A short list:

1. Check eligibility. Important pages indexed, snippets allowed, Search generative AI set to Include.
2. Stop anything that looks like fan-out spam. No pages per query variant.
3. Write one piece of non-commodity content a week instead of five generic ones.
4. Put a clear answer near the top of each page, then the depth. It helps readers first, and it's easier for any system to quote.
5. Check the Generative AI performance report monthly.

That's the list. It's mostly good SEO with a new report attached, which is what Google keeps saying. The same loop, with AI-visibility tracking across ChatGPT, Perplexity, Gemini and Claude, is what we're building into Autopilot, in early access now.