How to rank in AI Overviews

First, check whether your queries trigger one at all

A substantial share of queries never produce an AI Overview. Google shows one where it judges a synthesised answer more useful than a list, which skews towards informational and comparison queries and away from navigational and transactional ones.

Working on queries that never trigger an overview is the most common wasted effort on this surface, and it is trivially avoidable: check, at scale, before you plan anything. How to track AI Overviews covers building a series that means something.

1. Take your target query list and record which queries currently produce an overview.
2. Repeat the check weekly for a month. The surface is unstable, and a query that triggers once may not be a reliable target.
3. Drop the queries that never trigger. Work the ones that trigger consistently.
4. Note which competitors are being cited on the ones that do, because that is your realistic source set to displace.

Second, rank for the query

Overviews are built from Google's main index, drawing on pages already performing for the query. This is not a legacy concern to be modernised away: it is the prerequisite. A page that ranks nowhere is not in the candidate set.

Ranking first does not guarantee inclusion, and inclusion does not require ranking first. What the ranking gives you is candidacy. Being in the top handful of results is roughly the threshold at which inclusion becomes plausible.

Third, make the passage extractable

Answer in the opening passage of a section

Selection is passage-level, not page-level. An answer in the twelfth paragraph loses to a clearer one on a weaker page. Say the answer, then qualify it: that discipline is answer engine optimisation.

One question per section

A section covering three things cannot be pulled for any of them cleanly.

No context dependence

A passage beginning "as noted above" is unusable. Every passage should survive being read in isolation.

Tables where the content is genuinely tabular

Comparisons, specifications and tier structures extract well as tables and badly as prose.

Question-shaped headings

Headings in the words a buyer would use, not internal labels. This helps the matching pass find the right section.

Keep the page current

Comparison and pricing content is re-retrieved often, and a stale page loses to a current one. This is the least interesting item on the list and reliably one of the most effective.

Fourth, do not block the surface by accident

Two directives govern whether you can appear at all. Googlebot builds the index the overview is generated from. Google-Extended governs whether your content can be used to ground Google's AI surfaces.

Google-Extended is widely described as a training opt-out and set accordingly, which quietly removes the site from AI Overviews and from Gemini while leaving conventional rankings intact. Google documents what each of its crawlers does. It is the single highest-leverage line in a robots.txt file for Google's AI surfaces, and one of the most commonly misconfigured.

Fifth, report it honestly

Being cited in an overview is worth real brand exposure and considerably less traffic than the equivalent conventional ranking, because the overview answers the question on the results page. Reporting citation and traffic as though they move together misleads whoever reads the report.

Track them as separate lines. Expect citation up and clicks flat or down on the queries where you win, and say so in advance so nobody is surprised by their own success. Reporting AI search to a board covers how to frame that.

Frequently asked questions

How long does it take to appear in an AI Overview?

If the page already ranks well, restructuring the opening passage can change things within a few weeks. If it does not rank, the timeline is however long conventional SEO takes for that query, because ranking is the prerequisite.

Why did we disappear from an overview we were in?

Overviews are unstable. The same query can produce an overview one week and not the next, with a different source set. This is why point-in-time measurement is close to meaningless and tracking has to be continuous.

Does schema markup get us into overviews?

Not directly. Structured data helps a machine parse the page, which is worth doing, but it will not put you into a candidate set you were not ranking for. Schema is a hygiene factor here, not a lever. See how AI Overviews build an answer for what is.

Sources

1. Google Search Central, AI features and your website. Google's own guidance on how its AI surfaces select and use content.