How to Rank in AI Overviews: What Google Picks and Why

Check eligibility and the Search generative AI control before you judge the content. Those checks decide whether a page can be used at all.

Google's answer is short. There are no additional requirements to appear in AI Overviews or AI Mode, and no special optimizations are needed.

Eligibility is the Search baseline. A page must be indexed and eligible to show in Google Search with a snippet. Google adds no further technical requirement, but it does not promise that a page will be crawled, indexed or served.

Google's May 2025 Search Central post gives the same baseline in plainer terms. Googlebot must not be blocked, the page must return an HTTP 200 status, and the page must have indexable content.

Google's optimization guide, last updated July 10, 2026, adds one step. A site must be included in Search generative AI features in Search Console to be eligible for display in those features. Check that setting before any other change.

For the crawl and index baseline in practice, read the SEO fundamentals page. It covers the same Search basics that AI features depend on.

Google lists fundamentals for AI features. They include allowing crawling in robots.txt and by any CDN or hosting infrastructure, internal links, page experience, important content in text form, images and video, and structured data that matches the visible text.

Product and local businesses can use Merchant Center feeds and Google Business Profiles to make products and services visible in AI responses and in other Google Search results.

Google's guide says the factor that will likely matter most over the long run is content that people find unique, compelling and useful. It ranks that above the other suggestions in the guide.

Google calls content commodity when it rests on common knowledge that anyone could have written. Commodity content adds little unique insight for readers.

Google's AI systems look at a variety of sources, and a unique viewpoint that stands out helps a page in that comparison.

Test each page with one question: could anyone have written this from common knowledge? If the answer is yes, add something only you can supply, such as your own test results or a clear point of view.

AI Overviews appear only when Google's systems judge them additive to classic Search. They often do not trigger.

Plan around that gap. A query with no AI Overview gives a page no chance to be cited in one, so list the queries that trigger AI Overviews before you build a program around them.

Plan for the sub-questions behind a query, not for one exact query. Google's documents explain how the answer is built, and vendor studies show how unstable its sources are.

Google calls the method retrieval-augmented generation, or grounding. Its systems rely on core Search ranking to retrieve pages from the Search index, then read those pages to write the answer with links.

Google defines query fan-out as a set of concurrent, related queries that the model generates to fetch more relevant results. Its lawn example includes best herbicides for lawns, remove weeds without chemicals, and how to prevent weeds in lawn.

Google says both AI Overviews and AI Mode may use fan-out, issuing several related searches across subtopics and data sources. That lets Google show a wider set of links than classic search does.

The two features are not one system. Google says AI Mode and AI Overviews may use different models and techniques, so the responses and links they show will vary.

Gemini 3 became the default model for AI Overviews on January 27, 2026. Studies run before that date may have tested an earlier model.

Surfer, an SEO software vendor, ran 1,600 runs with Gemini grounding. Only about 27% of fan-out queries stayed consistent across repeated runs of the same prompt, and 66% appeared only once.

Surfer concludes that a fan-out list is a snapshot, not a keyword list you can rely on next week.

Semrush says no one can know in advance which fan-out queries Google will use, because they are generated dynamically and can vary with each search.

The clusters are steadier than the single queries. Surfer found that 84% of fan-out queries shared at least one top-20 URL with the original query, and that about 90% fit into four or fewer clusters.

In the same Surfer article, a test of 10,000 keywords and 33,000 generated fan-out queries found a Spearman correlation of 0.77. The more fan-out queries a page ranked for, the more likely it was cited in the AI Overview.

That result is a correlation. It shows a pattern in one vendor's sample, not a cause, and the pattern is not a rule for how you should build pages.

Google's guide sets a limit on the same idea. Making separate content for every variation or fan-out query, primarily to manipulate rankings or generative AI responses, violates Google's scaled content abuse spam policy.

Vendor advice points the other way. Surfer's fan-out article advises covering fan-out topics, so the vendor advice and Google's warning conflict on this point.

Resolve the conflict by covering the related sub-questions inside one useful page, or inside a real cluster of pages. Do not publish one thin page per variation.

Google says its AI systems understand relevance even when a query and a page's main content do not match exactly. You do not need to capture every long-tail variation.

Semrush ran a small test on four of its own articles, and citations in tracked prompts rose from two to five. Semrush calls the sample small and says AI citations can be unpredictable, so do not treat the result as proof.

Treat overlap figures as ranges from vendor studies, not as rules. Check each study's scope before you compare its numbers, because the samples and the features measured differ.

Ahrefs, a vendor, studied 863K keyword SERPs and 4M AI Overview URLs in a study published March 2, 2026. It found 37.9% of cited URLs also appeared within the first 10 SERP blocks, 31.2% at positions 11 to 100, and 31.0% beyond 100.

Ahrefs also counted blue links only, ignoring ads and SERP features. That split was 37.10% in the top 10, 26.20% at positions 11 to 100, and 36.70% outside the top 100.

Ahrefs' earlier study of 1.9 million AI Overview citations, published in July 2025, found about 76% in the top 10. Its March 2026 update puts that figure at about 38%.

Use the 38% figure as the current result. Ahrefs says its March 2026 update also improved its parsing method.

Ahrefs reads the drop as AI Overviews relying less on direct search results and more on sources in fan-out query SERPs. That reading is Ahrefs' inference, not a measured result.

Surfer, another vendor, studied 173,902 cited URLs. It found that 67.82% of cited pages did not rank in the top 10 for the query or for any fan-out query.

For the top three visible citations, that share fell to 45.86%. Most of the citations people see do rank, so put your effort into pages that can reach the top three.

Ahrefs found that 18.2% of cited pages that did not rank in the top 100 were YouTube URLs. YouTube was 5.6% of all AI Overview URLs cited in its dataset, so plan for video as a citation source in its own right.

seoClarity studied AI Mode, not AI Overviews. On 1,000 transactional US desktop queries from September 2025, it counted 12,011 AI Mode citations, and only 19% of them came from the top 20 organic results.

When seoClarity looked only at the top 10 rankings, overlap occurred for 28% of queries.Rank position one was cited in AI Mode results 25% of the time, rank two 21% and rank three 16%.

Do not apply AI Mode figures to AI Overviews. Google says the two may use different models and techniques, so their links can differ.

Ahrefs found near-zero correlation between word count and AI citations, with a Spearman value of about 0.04. Length alone is not a lever for citations.

Ahrefs' Head of Marketing Strategy & Research, Glen Allsopp, found that Google AI Overviews strongly favor "best" posts, with nearly 50% of AI Overview citations falling into this category. Treat it as a vendor finding.

In a study of about 17M citations across seven platforms, Ahrefs found that AI Overviews cite URLs that are 16 days older on average than organic results. It is the only platform in that study that skews older.

The top three AI Overview citations averaged over 1432 days since publication and 1047 days since their last update. Freshness is a reason to update the pages you want cited, not a reason to publish more pages.

Ahrefs compared 75,000 brands. Branded web mentions correlated 0.664 with AI Overview brand mentions, and backlinks correlated 0.218. Ahrefs states that correlation is not causation, which is the caveat to keep when you read these figures.

Ahrefs tracked 43,000 keywords that each had at least 16 AI Overviews over a month. Between observations, an AI Overview changed 70% of the time, and the average persistence of a cited source was 2.15 days.

Nearly half of the cited sources are new from one observation to the next, at 45.5%. Track citations across several observations, because one snapshot can mislead.

Google's guide names several tactics that it says you do not need. Where vendor advice points the other way, treat Google's statement as the baseline and the vendor claim as unproven.

llms.txt does not help. Google Search ignores it, so creating one neither helps nor harms your visibility or rankings.

Google's AI features page adds that you need no new machine-readable files, AI text files or markup to appear in these features.

Chunking is not required. Google says there is no requirement to break your content into tiny pieces for AI to better understand it, and there is no ideal page length.

You do not need to rewrite for AI. Google says you do not need to write in a specific way just for generative AI search.

Schema for AI is not required. Google says structured data is not required for generative AI search, and there is no special schema.org markup to add. Structured data still supports rich result eligibility, so keep it for that purpose.

Brand mentions are the sharpest disagreement. Google says that seeking inauthentic mentions across the web is not as helpful as it might seem, and that its core ranking systems favor high-quality content while other systems block spam.

A vendor correlation between mentions and citations does not show that the mentions cause the citations. Build mentions through real work, such as research people cite, not through placement counts.

Google also warns that no third-party tool has access to its internal ranking or AI systems, so be wary of any tool that claims to show them.

Pick the control that matches the goal: stop a passage, stop a page, stop AI features for a whole property, or stop training. Each control has its own scope, so a control set for one goal does not solve another.

When two robots rules conflict, the more restrictive rule applies. Nosnippet wins over a max-snippet value on the same page.

Robots meta tags are read only when a crawler can reach the page. A robots.txt block hides the tags, so a page blocked in robots.txt cannot pass its nosnippet to AI features.

Use the URL Inspection tool to confirm that Googlebot sees the control. Then allow time for a recrawl.

The SEO auditor checks crawlability and snippet tags, so run it on each URL you change.

The Search generative AI control sits under Settings, then Search generative AI, in Search Console. Include is the default. Exclude stops your links and content from those features, and Inherit takes the setting from the parent property.

Excluding does not change how your content ranks elsewhere. The control is not a ranking signal outside the generative AI features, and content from other sites still appears in them and may look similar to yours.

The control covers AI Overviews, AI Mode and generative AI features in Discover together. In Google's documentation, no control turns off AI Overviews alone while AI Mode stays on.

Google says content is excluded quickly after the control goes live, but caching and propagation across Google systems can delay some content.

By default, a property inherits the setting from its closest parent that has changed its control. A section such as /blog can carry its own setting when it is a separate property in Search Console.

The Search generative AI control does not stop AI training. Use Google-Extended to limit training, and noindex to block a page from Search entirely.

Google-Extended is a robots.txt token, not a separate crawler. Google says it does not affect a site's inclusion in Google Search and is not used as a ranking signal there.

Preferred Sources is a choice that readers make. Google says people can choose the websites they want to see more prominently in AI Overviews and AI Mode, so the announcement describes a reader setting, not a site setting.

Set a baseline in the Generative AI report before you change any control. The report counts impressions only, so pair it with clicks from the Performance report for the same pages.

Google launched its search generative AI performance reports on June 3, 2026. By August 31, 2026, Google had rolled them out to all websites worldwide.

The report shows impressions only. An impression counts how often URLs from your site appeared in generative AI features in Search and Discover. Google says it will add metrics over time, but the report lists no clicks, click-through rate or position today.

You can group the impressions by page, country and device, and by date at hourly, daily, weekly or monthly level.

The report covers AI Overviews and AI Mode together, and Google does not list a split between them. Search Console leaves out experiments in Search Labs.

The chart counts two results from one site as one impression. The table groups by page, so the chart total and the table total can differ.

The report uses the same data limits as the Search Performance report, including the 1,000 row limit. Plan your exports around that limit from the start.

A report can be missing. Google says this usually happens when the site has too few impressions in generative AI features, or when the control excludes the site.

Clicks from AI features stay in the Performance report, within the Web search type.The generative AI data is also part of the overall Performance report, so do not add its impressions to the Performance totals.

All links in one AI Overview share one position. Clicking an external link in an AI Overview or AI Mode counts as a click.

A follow-up question in AI Mode counts as a new query, so one conversation can add several query rows.

Third-party tools sample these results. Ahrefs notes that Google does not expose the simulated fan-out queries it generates for each AI response, so treat any tool's fan-out list as a sample.

Set the value of a citation against the click loss on your own queries, not against one industry figure. The studies differ in sample and design, so compare them with care.

Ahrefs found AI Overviews on 20.5% of desktop SERPs in a September 2025 sample of about 146 million desktop SERPs. Of the keywords that triggered an AI Overview, 99.9% were informational.

Pew Research Center measured clicks in a panel of 900 US adults. Users who saw an AI summary clicked a traditional result in 8% of visits, against 15% of visits without one.

Pew also found that users clicked a link inside the AI summary in 1% of visits.

Ahrefs found that AI Overview presence correlates with a 58% lower average click-through rate for the top-ranking page. The study compared 300,000 keywords in two groups, using Search Console desktop data from December 2023 and December 2025.

Seer Interactive measured informational queries across 2025. Organic CTR was 2.07% when the brand was cited in the AI Overview, 0.94% when an AI Overview showed and the brand was not cited, and 3.35% with no AI Overview.

Seer says it cannot claim causation. It also says one account, 47% of that segment's impressions at 0.20% CTR, pulls the 0.94% down, and that without that account the figure is approximately 1.61%. Treat the gap as a reason to measure your own queries, not as a forecast.

Google's own statement covers a different measure. It says total organic click volume to websites has been relatively stable year over year, and it gives no figures.

Google reports total click volume. The studies report click-through on single queries when an AI Overview shows. Those are different measures, so do not set one against the other.

Google also says clicks from AI Overviews are higher quality, meaning users are more likely to spend more time on the site. It gives no number for that claim.

Google urges site owners to look at conversions, such as sales, signups, a more engaged audience or information lookups about the business, not only at clicks.

Google says AI Overviews send people to a greater diversity of sites for more complex questions. It gives no figure for that claim.

Google also reports user scale. AI Overviews have over 2.5 billion monthly active users, and AI Mode has surpassed one billion monthly users.

Judge a citation by the conversions it brings from the queries that show an AI Overview. Use the click-through figures above as a range, not as a target.

Work through these five steps in order. Each step depends on the one before it.

1. Check eligibility: the page is indexed and eligible to show a snippet, robots.txt does not block Googlebot, and the Search generative AI control is set to Include.
2. Check fan-out coverage inside real pages. Cover each sub-question in a useful page or a real cluster, and do not publish one thin page per variation.
3. Decide the controls. Use the snippet rules for passages and pages, the Search generative AI control for a whole property, and Google-Extended for training.
4. Set a measurement baseline. Record impressions by page from the Generative AI report, and clicks and conversions for the same pages from Performance and your analytics.
5. Ignore the tactics that Google says you can skip: llms.txt, chunking, rewrites for AI, AI-specific schema and inauthentic mentions.

Record the baseline again after each change, because citations move between observations. Run the complete SEO checklist on each changed page, and read the search ecosystems page for platforms other than Google.