Google AI Overviews: how to rank in them and what you control

Google AI Overviews are the AI-written answers at the top of many Google searches, and the way to rank in them is to rank in Google Search: Google builds AI Overviews from its normal index and ranking systems, and its only documented requirements are that a page is indexed, eligible to show a snippet, and on a site that has not opted out in Search Console. There is no special markup, file or "AI version" of a page. Below you will find how Google picks the pages it links, how to optimize your content for AI Overviews, the controls you have if you would rather stay out, and a practical routine for tracking your visibility with Search Console and a little manual sampling. It is part of our guide to generative engine optimization, which covers AI search as a whole.

Quick answer

• AI Overviews use Google's search index. If a page is not indexed or cannot show a snippet, it cannot appear. Nothing else is required, as long as the site has not opted out.
• No special markup or files. Google's guide to its generative AI features says you do not need AI text files, special markup or Markdown versions of pages, and that llms.txt gets no special treatment.
• Pages are found through "query fan-out": Google may run several related searches for one question, so a page that answers a specific sub-question well can be linked even if it does not rank for the main query.
• Write non-commodity content: your prices, your data, your experience. Google warns that generic pages repeating what everyone says are not what it wants.
• You can opt out: the Search generative AI control in Search Console removes a site from AI Overviews and Google's other generative search features; nosnippet, data-nosnippet and max-snippet limit what can be quoted. Google-Extended does not affect AI Overviews.
• Track it in Search Console's Generative AI performance report (impressions only, by page, country, device and date), plus a monthly hand check of 10 to 20 queries. Clicks from AI Overviews exist, but they hide inside the Web totals of the regular Performance report.

What are Google AI Overviews?

AI Overviews are the AI-written summaries shown above or among Google's normal results for some searches, with links to the pages they draw on. They sit inside the ordinary results page, so they reach everyone who searches, not only people who choose an AI product. Since Google I/O in May 2026, people can ask a follow-up question straight from an AI Overview and carry on in a conversation in AI Mode, Google's fully conversational search mode. For site owners, both features share the same eligibility, the same controls and the same Search Console report.

The links shown with an answer are the visibility you can get: a place in the answer, with a clickable link to your page next to the text it supports. Google keeps changing how those links look. On 15 October 2026 it announced more inline links placed higher in AI Overview and AI Mode answers, a more clickable "Dive deeper" link section with image thumbnails, and a test of a dedicated section for official websites at the top of some AI Mode answers.

How Google picks the pages it links

Google's documentation of its AI features names two techniques:

1. Retrieval-augmented generation. The model does not answer from memory alone. It retrieves pages from the Search index and grounds its answer in them, which is why only indexed pages can appear.
2. Query fan-out. For a complex question, AI Overviews and AI Mode may run several related searches across subtopics. Google's example is a question about fixing a lawn full of weeds, which might fan out into searches such as the best herbicides for lawns, removing weeds without chemicals and how to prevent weeds. Results from all of those feed the answer.

Two practical consequences follow. First, the pages used are chosen by Google's core ranking and quality systems, so everything that helps a page rank helps it here. Second, fan-out means you can be linked for a question you would never rank for on its own, because one of your pages answers one of its parts better than anyone else. It also means the reverse: a page that ranks first for the exact phrase someone typed is not guaranteed a link, because the answer is assembled from the fan-out searches.

What Google documents, and what it says you do not need

Google published "Optimizing your website for generative AI features on Google Search" on 15 May 2026. Its position is that optimising for AI Overviews and AI Mode is "still SEO": the generative features are rooted in Google's core Search ranking and quality systems, so the same work serves both. The guide has a mythbusting section on popular "AEO" and "GEO" advice.

What the guide does ask for is familiar: crawlable, indexable pages; helpful, people-first content with a distinctive point of view; good images and video; a sound page experience; and, for shops and local businesses, up-to-date data in Merchant Center and Google Business Profile, which feed shopping and local answers. Google also states that its spam policies apply to generative AI responses in Search.

How to optimize content for Google AI Overviews

The steps are ordered by how often they are the real problem: most pages missing from AI Overviews fail on eligibility before they fail on content.

1. Make sure the page is indexed

Search Console, URL Inspection, tells you whether a URL is on Google. If it is not, fix that first; why a page is not in Google walks through the causes. The report checks the basics on every page:

2. Check that nothing blocks snippets

A page with <meta name="robots" content="nosnippet">, or a max-snippet:0 directive, cannot be quoted in AI Overviews. These tags often come from an SEO plugin setting or a template that was meant for one page type. Search your templates for nosnippet and max-snippet, and check the X-Robots-Tag response header too, because a header applies even when the HTML looks clean. The test in How we tested shows a two-line way to check both.

3. Put the content in the HTML

Googlebot renders JavaScript, so AI Overviews can use content that scripts add. Rendering can still be delayed or incomplete, though, and other AI search products do not render at all. Main text, prices and answers belong in the server's HTML:

4. Answer the sub-questions, not just the head term

Think about how your topic fans out. A page about a hand-made leather bag can answer, in plain sentences near the top of each section: what leather it is, how long it takes to make, the price, delivery times and costs to your main markets, and how to care for it. Each of those is a sub-question someone's search might fan out into. Put the answer first under a clear heading, then the detail.

5. Add what only you can say

Google's guide warns against "commodity" content. A generic "10 tips" article loses to a page with first-hand detail: measured numbers, your own photos, real prices, what went wrong when you tried it. That is also what makes a page worth citing.

6. Keep your feeds and profiles current

For products, keep Merchant Center feeds accurate: price, availability, shipping. For a local business, keep Google Business Profile hours, address and categories right. These feed shopping and local answers directly.

7. Structured data: useful, not required

Structured data does not make a page eligible for AI features, and Google says no special markup is needed. It remains useful for rich results and for stating facts such as price and availability precisely; if you add it, it must match what is visible on the page. Our guide to schema markup for AI search looks at the evidence per operator.

What you control if you want to stay out

The Search generative AI control arrived with the new reports in June 2026, first for a subset of sites, and reached all sites worldwide on 31 August 2026. Every property starts on Include; you switch it to Exclude to leave. Google says opting out does not affect your ranking in the regular results, so your blue links stay; you lose the impressions and traffic from AI features. It covers Google Search's generative features only, not the separate Gemini app or model training.

The snippet directives are blunter. nosnippet also removes your normal text snippet, which usually costs clicks, so prefer data-nosnippet on the specific parts you do not want quoted.

How to track AI Overviews: Search Console, sampling and their limits

No single source shows the whole picture. Search Console counts how often you appeared, a hand check shows for which questions, and analytics shows visits without telling you which came from an AI Overview. Here is what each source answers:

Step 1: read the Generative AI performance report

In Search Console, open Performance, then Generative AI. Google announced the report on 3 June 2026 and it reached all sites worldwide on 31 August 2026. It counts impressions: how many times links to your site were shown in AI Overviews or AI Mode, with views by page, country, device and date. Discover's generative features have a separate view. Small sites may see no data until they have enough AI impressions, and a site that opted out has nothing to show.

What it cannot show: queries, clicks, click-through rate and position are absent. You can see that a page appeared in AI answers 4,000 times last month, but not for which questions, and not whether anyone clicked.

A useful monthly routine with it:

1. Set the date range to the last full month and open the breakdown by page.
2. Export the list, and keep each month's file, so you build your own history.
3. Compare it with the pages you want to appear. A page you care about with zero impressions is your first eligibility check (URL Inspection, snippet directives).
4. Watch the trend per page for four to six weeks after you change one, rather than the site total, which hides what moved.

Step 2: find the clicks in the regular Performance report

Google states that clicks and impressions from AI Overviews and AI Mode are included in the overall Web search type totals of the Performance report, and there is no filter to separate them. A page's "Web" clicks therefore mix classic blue-link clicks with AI feature clicks, which is why a page can gain impressions in the Generative AI report while its click curve looks unchanged. Put the two side by side per page: rising AI impressions with flat clicks is common, and it tells you the answer is being shown, not that the page is broken.

The Performance report also gives you the raw material for step 3: the queries that already bring a page impressions are the best candidates to sample by hand.

Step 3: sample the queries by hand once a month

Pick 10 to 20 queries that matter to you: the top queries of the pages in the Generative AI report, plus the questions your customers actually ask. Once a month, search each one and record a row per query in a spreadsheet:

• date, query, device and country;
• whether an AI Overview appeared;
• whether your site was linked, which page, and whether the link was inline in the text or in the list of links;
• which other sites were linked.

Search signed out, in a private window, on the same device each time, and repeat a query once if the answer looks unusual, because AI Overviews vary between runs, places and people. Twenty rows a month will not give you a precise share, but over three months it shows whether you are gaining or losing the questions you care about and which competitors take your place. Third-party AI overview trackers automate the same idea on a larger list, with the same limits: they see only the queries you give them, their answers vary by location and run, and none has Google's impression data. Treat any tracker's numbers as trends on a sample, not as your visibility.

Step 4: know what analytics and logs cannot show

A visitor who clicks a link in an AI Overview arrives as an ordinary Google organic visit, from a google.com referrer, so GA4 and other analytics tools cannot separate it from a blue-link click. Our AI referral tracking channel group counts visits from assistants such as ChatGPT, Perplexity, Claude, Gemini and Copilot, which do send their own referrers, but by design it cannot pick out AI Overviews.

Server logs cannot help either. Google has no separate crawler for AI Overviews: pages are fetched by Googlebot for the normal index, and Google-Extended is a robots.txt token, not a user agent you will ever see in a log. Logs are still worth reading for the other AI crawlers; detecting AI crawlers in server logs shows how. A dedicated guide to tracking AI search visibility across engines is coming; until then, Search Console vs getReport explains which tool answers which question.

How to check your page is eligible

To check the technical side in one go, the AI crawler check shows whether your content is readable without JavaScript and what your robots.txt says to Google-Extended and other AI tokens:

A passing result for AI Overviews looks like this: the page is indexable, no nosnippet or max-snippet:0 appears in the HTML or the X-Robots-Tag header, the main text is present without JavaScript, and Search Console's Generative AI setting reads Include. Remember that blocking Google-Extended in the check's results is fine for AI Overviews; it does not remove you. AI search visibility covers the measurement sources beyond Google.

Common mistakes

• Blocking Google-Extended to leave AI Overviews. It has no effect on them. Use the Search generative AI control or snippet directives.
• A leftover nosnippet or max-snippet:0. The page ranks but can never be quoted.
• Mass-producing question pages. One page per phrasing of a question is what Google calls scaled content abuse.
• Adding schema and expecting AI Overviews. Structured data does not create eligibility; indexing and snippet eligibility do.
• Judging success by clicks alone. The Generative AI report shows impressions only, and the clicks that do happen hide inside the Web totals; watch impressions per page as well as traffic.
• Reading one search as the truth. A single AI Overview you saw once is a sample of one; record a monthly sample before you change a page because of it.

How we tested

The getReport team re-checked this guide on 16 October 2026. We compared every claim about eligibility, controls and reporting with Google's AI features documentation, its May 2026 guide to generative AI features, the Search Console help for the Generative AI report and Google's announcements up to 15 October 2026, and removed or corrected anything we could not match to them.

We ran the snippet check from step 2 with curl 8.5 against two getReport fixtures served locally: our fixture of the example-shop.hr demo shop on port 4101, and the noindex-meta fixture page on port 4100. The command reads the X-Robots-Tag header and the robots meta tags, and looks for data-nosnippet:

for u in http://localhost:4101/ http://localhost:4100/pages/noindex-meta/; do
  echo "== $u"
  curl -sI "$u" | grep -i '^x-robots-tag' || echo "x-robots-tag: (none)"
  curl -s "$u" | grep -oiE '<meta[^>]+name="(robots|googlebot)"[^>]*>|data-nosnippet' || echo "robots meta: (none)"
done

Real output:

== http://localhost:4101/
x-robots-tag: (none)
robots meta: (none)
== http://localhost:4100/pages/noindex-meta/
x-robots-tag: noindex, nofollow
<meta name="robots" content="noindex, nofollow">

The shop home page sends no restriction in either place, so nothing stops it from being quoted. The second page is blocked twice, once in the header and once in the HTML, which is why checking only the HTML is not enough. Run the same two lines against your own URLs and look for noindex, nosnippet or max-snippet:0.

What we did not test: we did not query Google, record AI Overviews or measure how often any page is linked. We did not switch the Search generative AI control on a live property, and the Search Console steps follow Google's help pages rather than screenshots of our own account.

Questions people ask

Do I need to optimise separately for Google AI Overviews?

No. Google says AI Overviews have no requirements beyond normal Search: the page must be indexed and eligible to show a snippet, and the site must not have opted out in Search Console. There is no special markup or file to add, and llms.txt gets no special treatment. Good SEO is the optimisation, plus answering specific questions clearly, since Google's query fan-out looks for pages that answer each part of a question.

How does Google choose which sites appear in AI Overviews?

Google retrieves pages from its Search index using its core ranking and quality systems, often running several related searches for one question, a technique it calls query fan-out. The model then writes an answer grounded in those pages and links some of them as sources. A page that answers one part of a complex question well can be linked even if it does not rank for the whole query.

Why is my site not showing in AI Overviews?

Check eligibility first: the page must be indexed (URL Inspection shows this) and able to show a snippet, so look for nosnippet or max-snippet:0 in your templates and headers, and make sure nobody set the Search generative AI control to Exclude. After that it is ranking: AI Overviews link pages that rank for the question or one of its fan-out searches, so a page that ranks for nothing will not be picked.

Can I stop my site appearing in AI Overviews?

Yes. Search Console's Search generative AI control, under Settings, removes the whole site from AI Overviews, AI Mode and generative AI features in Discover, and Google says it does not affect your ranking in normal results. For individual pages or passages, use nosnippet, data-nosnippet or max-snippet. Blocking Google-Extended in robots.txt does not remove you.

Does structured data help you get into AI Overviews?

Not directly. Google's guide says no special markup is required for its generative AI features, and adding it does not make a page eligible. It still helps with rich results and states facts such as price and availability precisely, which is useful to any system reading the page. If you add it, keep it valid and matching what is visible on the page.

Where do I see AI Overview impressions for my site?

In Search Console, under Performance, then Generative AI, available to all sites worldwide since 31 August 2026. It shows how often links to your pages appeared in AI Overviews and AI Mode, by page, country, device and date; Discover's generative features have their own view. It shows no clicks and no queries, so pair it with a monthly manual sample of your key queries.

Can Google Analytics show traffic from AI Overviews?

No. A click on a link in an AI Overview arrives as a normal Google organic visit with a google.com referrer, so GA4 cannot tell it apart from a blue-link click. Search Console counts those clicks, but inside the Web totals with no filter. Assistants such as ChatGPT, Perplexity and Gemini do send their own referrers, so a custom channel group can count those.

Sources

• AI features and your website, Google Search Central (accessed October 2026)
• Optimizing your website for generative AI features on Google Search, Google Search Central (published 15 May 2026, accessed October 2026)
• Introducing Search Generative AI performance reports in Search Console, Google Search Central Blog (3 June 2026, updated 31 August 2026)
• Generative AI performance report (Search), Search Console Help (accessed October 2026)
• Google's common crawlers: Google-Extended, Google Search Central (accessed October 2026)
• AI Mode and AI Overviews updates to explore the web, Google (15 October 2026)
• Google Search's I/O 2026 updates, Google (May 2026)