How to rank in Google's AI Overviews (what actually works)

TLDR; There is no separate submission or special schema for AI Overviews. Google says the same SEO fundamentals apply: make important content crawlable, internally discoverable, visible as text, technically sound, and useful. Clear answers and supporting evidence are sensible editorial practice, but Google does not publish a guaranteed format or classic-ranking threshold for being cited.

AI Overviews appear for a material but changing share of informational queries. The rate varies by topic, market, device, and study period. In a recent sweep of my own target terms, the answer box appeared on nearly all of them, but that small, selected sample is not a market-wide prevalence estimate. The planning implication is narrower: measure the queries that matter to your site rather than importing a global percentage.

So the question clients ask, “how do we rank in that?”, needs reframing. Google composes an answer using its search systems and may cite supporting pages. There is no separately submitted AI Overview ranking. Here is the practical work I use, separated from what Google officially documents.

You don’t rank, you get extracted

A classic result is a link Google ranks so a human can click it. An AI Overview is an answer Google writes, drawing on sources it can lift a clean passage from and cite. Those are different mechanisms, and the second one rewards content built to be quoted, not just found.

You are not trying to be the best page. You are trying to be the best extractable answer to a specific question, on a page Google already trusts.

What Google pulls into an AI Overview

The following traits are observational editorial patterns, not published Google eligibility rules:

• They already rank for the query or close variants. The Overview is not a way around search authority; it draws on pages that have it.
• They contain a direct answer to the exact question, near a heading that matches it, rather than an answer buried in a long preamble.
• The answer is self-contained: it makes sense read on its own, without the surrounding paragraphs.
• They carry specifics and evidence, numbers, steps, a clear position, over generic summary.
• They come from a credible, identifiable source, a real author and site the engine has reason to trust on the topic.

The practical checklist

This is the work, in order of leverage:

1. Mine the real questions. Pull them from People Also Ask, Search Console and your own sales conversations. These are the extraction targets.
2. Match a heading to each question. Phrase the heading the way people ask, then answer it in the first sentence or two below.
3. Lead with a concise answer, then expand. Give the self-contained response first; add nuance and detail after. A buried answer does not get lifted.
4. Add supporting evidence. Data, a worked example, a specific recommendation. This is what separates a citable source from a skippable one.
5. Earn the underlying ranking. Topical depth, internal linking, and the technical health that conventional SEO is built on. Without it, the rest rarely gets you cited.
6. Add question-and-answer structure where it helps the reader. Use matching structured data only when it is supported for the page type and accurately reflects visible content. Google explicitly says there is no special schema required for AI Overviews, and FAQ rich results ended on May 7, 2026, so do not add FAQ schema as AI Overview insurance.

Can you opt out or beat them?

You cannot selectively opt out of AI Overviews while staying in normal search: Google uses the same index for both. Blunt controls like nosnippet or max-snippet limit the text Google can show, but they cut your classic visibility too, so they are rarely worth it for content you want found.

“Beating” an Overview is the wrong frame. When the Overview appears and you are cited in it, you stay visible even as it absorbs the click. When no Overview appears, strong conventional ranking still wins the query. Either way the answer is to do both, which is why this sits on top of SEO rather than replacing it.

How to measure it

Track three things:

1. Whether your target queries show an AI Overview.
2. Whether your site is cited when they do.
3. What happens to the affected pages over time.

Third-party SERP tools can record Overview presence and citations. Google includes AI-feature traffic in Search Console’s Web performance reporting and has offered a generative-AI performance report to all sites since August 31, 2026, so use the reports available in your account and document their scope. Assistant citations across ChatGPT, Claude and Perplexity still need separate monitoring because they are different products and indexes.

The UK online training provider case study uses Search Console for its result: in July–September 2026 its UK organic clicks reached 4.6x the same quarter of 2025, though the figure does not separate AI Overview clicks from classic ones.

For the wider framing of how this fits with search, see AEO vs SEO: what answer engine optimisation actually changes. If you want your own content audited for where it triggers AI Overviews and where it is being passed over, that is part of my Organic Growth and AI Search work, which also connects the visibility you earn to enquiries you can trace.

Sources

• AI features and your website, Google Search Central. Eligibility, SEO fundamentals, measurement and content controls for AI Overviews and AI Mode. Primary documentation
• FAQ rich results in Search Console, Google Search Console Help. Records the May 2026 removal of FAQ rich-result reporting after FAQ rich results stopped appearing. Primary documentation
• Generative AI performance report, Google Search Console Help. Scope and availability of Google’s dedicated generative-AI reporting. Primary documentation