How to optimize for Google AI Overviews
Google AI Overviews are the AI-written answers that now sit above the classic results, and getting cited in them is less mysterious than the industry pretends: 38% of AI Overview citations come from pages already ranking in the top 10, and the rest skews heavily toward pages ranking somewhere. Optimizing for AI Overviews is therefore ranking work plus extraction work, and this guide covers both halves with their evidence.
How does Google build an AI Overview?
Through query fan-out. Google’s own AI Mode documentation describes the mechanic: your question is broken into subtopics, a batch of related searches runs simultaneously, and the model synthesizes an answer from what those searches return, citing sources as it goes. Two practical consequences follow. Ranking remains the entry ticket, since the fan-out searches pull from ordinary results.
And the queries being matched are not the user’s literal words but generated variants, often narrower and often carrying modifiers like “best,” “cost,” or the current year, which is why covering the specific questions inside your topic beats one broad page.
Which pages earn the citations?
Pages that rank for a fan-out query and make extraction cheap. Once retrieved, your page competes at passage level: the model wants a self-contained chunk that answers the sub-question cleanly, and it finds those chunks disproportionately at the top of pages, where Surfer measured almost 40% of AI citations originating in the first 100 words.
A page that opens with its claim, then answers each sub-question under its own question-phrased heading, presents the model a menu of liftable passages. A page that buries its point after 600 words of scene-setting usually donates the citation to someone terser.
What on-page work should you do first?
In order of leverage. Confirm retrievability, because a firewall silently blocking crawlers or content that only exists after JavaScript removes you from consideration regardless of rankings; our scanner tests both, and the firewall and JavaScript guides fix them. Restructure key pages answer-first, with the point in the first 100 words and question headings matched to real customer phrasing.
Then expand coverage of the sub-questions in your topic, since each one is a fan-out query you can rank for. Titles matter more than they look: pages get judged by title, snippet, and URL before being opened, per Ahrefs’ citation research.
Does schema markup get you into AI Overviews?
The evidence says no. Ahrefs tracked 1,885 pages that added schema and found no increase in AI citations, including in AI Overviews specifically, and a peer-reviewed controlled study found LocalBusiness schema moved nothing on Google surfaces. Schema keeps its classic value for rich results, which can lift click-through on the classic listings, so add the basic version.
But if an agency’s AI Overviews pitch is built on structured data, the measured record is not on their side.
How should you measure AI Overview visibility?
Track rankings for question-shaped queries first, because they lead everything else, and then spot-check the Overviews themselves monthly for your most valuable questions. Google’s results vary by user and session, so treat individual sightings as anecdotes and the trend as the signal.
Search Console shows impressions that include AI Overview appearances without isolating them, which is imperfect but directionally useful when a page’s impressions jump after an answer-first rewrite. For the full generative-search picture beyond Google, the GEO guide covers ChatGPT search and Perplexity, and the free scan measures whether your pages are ready for any of them.
See where your site stands. The free scan takes about fifteen seconds and shows every fix.
Scan your site for AI readiness