Quotable answer: Google AI Overviews pull an answer together from several sources at once, then decide which ones to name. The clearest public signal is a technique Google calls query fan-out: instead of answering only the exact question typed, the system runs several related searches behind the scenes and pulls passages from whichever pages answer each sub-question most directly. A page does not need special markup to qualify, only to already be indexed and eligible for a normal search snippet. Beyond that baseline, the pattern that correlates with getting named is corroboration: content that states a claim clearly, in language other trusted pages also use to describe it, and that a search engine can lift as a self-contained passage without extra context. Google has not published the full ranking formula, so anything more specific than this is informed pattern-reading, not confirmed mechanics.
A founder asks us some version of this every month: "Why did my competitor get named in the AI Overview and I didn't? Our page ranks higher." It's a fair question, and the honest answer is that ranking and getting cited by an AI Overview are no longer the same contest. What's sometimes called answer engine optimization, the practice of writing for the systems that answer questions directly instead of just listing links, is really an attempt to reverse-engineer that second contest. Here's what's actually known about how it works, and where the certainty stops.
What Google actually says about how it picks sources
Google's source-selection process uses a technique it calls query fan-out. Instead of matching a page to the one query someone typed, the system breaks that query into several related sub-questions, searches each one separately, and pulls the clearest passage for each from wherever it lives. Google's own documentation for AI features puts the eligibility bar simply: "a page must be indexed and eligible to be shown in Google Search with a snippet," meeting the same technical requirements as ordinary search results. There is no extra tier to unlock.
Two lines from that documentation matter more than the rest of it combined. First: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." Second: no special schema.org markup is required to be eligible. That contradicts a lot of what gets sold as "AI SEO." A page cannot buy its way into an AI Overview with a markup trick. It gets there by being the clearest, most retrievable answer to one of the sub-questions the system generated on its own.
That single fact reframes the whole exercise. A brand isn't optimizing for one search box. It's trying to be the best answer to several questions it never sees, generated automatically from the one a reader actually typed.
The signals that correlate with getting named
None of these are confirmed ranking factors in the way classic SEO factors were once documented. They're the pattern that repeats across the pages that keep showing up in AI Overviews, and they line up with what Google's own documentation implies about how the system works.
- Corroboration. A claim that several independent, trusted pages state the same way gives the system more confidence to lift it. A brand saying something nobody else says, in language nobody else uses, is harder to verify against other sources, even when it's true.
- Structured, quotable phrasing. Short, direct sentences that answer one question and stand alone, without needing the paragraph before or after to make sense, are what a passage-extraction system is built to grab.
- Entity clarity. One consistent brand name and description, repeated the same way across the site, third-party profiles, and press mentions, is easier for a system to resolve to a single, trustworthy source.
- Freshness. A page that reflects the current state of a fast-moving topic, with a visible last-updated date, is a safer pick than a page that reads like it hasn't been touched since the facts changed.
- Being cited elsewhere. Pages that other trusted sites already reference, link to, or quote give the system a second, independent signal that the source is worth citing again.
The one number worth trusting
Most "AI Overview ranking factor" content online cites numbers nobody can trace back to a named study. One that holds up: Ahrefs analysed 863,000 keyword SERPs and 4 million AI Overview URLs through its Brand Radar tool and published the finding on 2 March 2026. Only 38% of pages cited in AI Overviews also ranked in the top 10 of the classic organic results for that same query, down from roughly 76% in an earlier version of the same study.
Read that number the way the researchers did: it isn't proof that ranking stopped mattering. It's evidence that fan-out increasingly pulls citations from pages that never appeared in the direct search results at all, because they answered one of the system's generated sub-questions instead of the original one. A page can rank nowhere for a brand's target keyword and still get cited, if it happens to be the clearest answer to a sub-question the system asked on its own.
What's known vs. what's speculated
Being honest about this split matters more than sounding confident. What Google has stated directly: the query fan-out mechanism exists, no special markup is required, and standard indexing eligibility is the technical floor. What's pattern-inference from watching thousands of citations, not confirmed mechanics: the relative weight of corroboration versus freshness versus entity clarity, whether schema markup gives any edge despite Google saying it isn't required, and exactly how much weight classic ranking signals like backlinks still carry inside the fan-out process. Content marketed as "how to get cited by ChatGPT" or a definitive checklist for "what is answer engine optimization" often states these unconfirmed patterns as settled rules. Some of it is a decent read of the pattern. None of it is Google's own account of the mechanism, because Google hasn't published one.
The one thing that isn't speculation: SparkToro and Similarweb's June 2026 study found that 68% of US Google searches now end without a click, up from roughly 60% two years earlier. That study is covered in full, along with what a brand does about it, in the zero-click survival plan. It's the reason this question, how a brand gets named inside the answer instead of just ranking under it, stopped being a niche concern for AI companies and started being a survival question for everyone else.
What this means for a smaller brand
A brand that can't out-publish a media site on volume still has a real path here, because the game isn't about the most pages. It's about being the clearest, most corroborated, most retrievable answer to a specific sub-question inside your category. That favours a smaller, sharper operation over a larger, vaguer one.
The practical version: write the direct answer first, before the backstory. State claims the way other trusted sources state them, don't invent your own vocabulary for a well-understood idea. Keep the brand name and description identical everywhere it appears. Update pages when the facts change and say so. None of that requires a markup trick or a growth hack. It requires the same discipline that used to define good writing before anyone called it AEO. We build this into every SEO and AI discoverability engagement, structuring pages so a system can lift the right passage cleanly, not gaming a formula nobody outside Google has actually seen. If that's the gap, our SEO and AI discoverability work is where we'd start.
FAQ
Does adding FAQ schema get a page cited in AI Overviews? Not by itself. Google's own documentation says no special schema is required for eligibility. FAQPage markup still helps some AI systems extract question-and-answer pairs cleanly, but it isn't a shortcut past the underlying content quality bar.
Is ranking #1 in Google still necessary to get cited? No, not always. Ahrefs' March 2026 analysis found only 38% of AI Overview citations also ranked in the classic top 10 for that query, down from about 76% earlier, because fan-out pulls citations from pages answering related sub-questions, not only the original one.
What is query fan-out? It's the technique Google says AI Overviews and AI Mode use: breaking one search into several related sub-questions, searching each separately, and pulling the clearest answer for each from across the web, rather than answering only the literal query typed.
Does a brand need a large content library to get cited? No. The signal that correlates with citation is being the clearest, most corroborated answer to one specific sub-question, not total page volume. A focused, well-corroborated page can outperform a large site with vaguer coverage of the same topic.
Can backlinks still help a page get cited in an AI Overview? Likely, but Google hasn't confirmed the exact weight. Pages other trusted sites already reference give the system an independent signal worth citing again, which lines up with how backlinks worked in classic ranking, but this specific correlation is pattern-inference, not a confirmed factor.
How is this different from classic SEO? Classic SEO optimizes for one ranking position against one query. AI Overview visibility means being retrievable as the answer to several sub-questions the system generates on its own, some of which never surface in a normal search results page at all.
Nobody outside Google has the full formula, and anyone selling a guaranteed AI Overview citation is selling something Google's own documentation contradicts. What's provable is that the mechanism runs on breadth of sub-questions answered clearly, not a trick applied to one page. That's a content and clarity problem, not a technical one, and it rewards a brand willing to do the unglamorous work of saying things plainly and backing them up.