Google AI Overviews: Source Eligibility, Visibility, and Measurement

A page can appear as a supporting link in Google’s AI Overviews only if it is indexed and eligible to be shown in Google Search with a snippet. Google says there are no additional technical requirements, no special markup and no AI-specific file to add. Eligibility is therefore an ordinary Search question; visibility is decided by Google’s systems query by query; and measurement, since 2026, has its own Search Console report that counts impressions but does not show clicks or queries. Each of those three layers answers a different question, and mixing them up leads to the wrong fix.

What AI Overviews and AI Mode are

Google describes AI Overviews as a way to help people “get to the gist of a complicated topic or question more quickly”, with links to explore further, and says they are shown only “when our systems determine that it is additive to classic Search” (AI features and your website). Google’s own page adds that, as a result, they “often don’t trigger”.

AI Mode is a separate experience in Search. Google describes it as useful for queries that need “further exploration, reasoning, or complex comparisons”, producing an AI-powered response with links to supporting websites. Both features may use what Google calls a “query fan-out” technique, issuing related searches across subtopics to build a response, which lets them display a wider set of supporting links than a classic results page.

The practical distinction for site owners:

Layer 1: eligibility

Google’s AI features guide states that to be eligible as a supporting link in AI Overviews or AI Mode, “a page must be indexed and eligible to be shown in Google Search with a snippet”, and adds that there are no additional technical requirements. The same page says site owners do not need to create new machine-readable files, AI text files or special schema.org markup to appear, which is why GEO and AEO labels do not name a separate Google system.

That gives a short eligibility checklist, every item of which is an ordinary Search check:

1. The URL is crawlable (not blocked for Googlebot in robots.txt).
2. The URL is indexed (URL Inspection shows it on Google).
3. The page is snippet-eligible: no nosnippet, and any max-snippet value is not set so low that it removes the text Google could use.
4. The site has not excluded itself through the Search generative AI control (below).
5. The page complies with Google’s spam policies.

Eligibility is a floor. Google’s guide also states that meeting all requirements and best practices “doesn’t mean that Google will crawl, index, or serve its content”. No method guarantees that a given page will be cited for a given query, and any service promising guaranteed AI Overview citations is promising something Google itself does not.

Layer 2: visibility

Once a page is eligible, whether it appears depends on the query, on whether an AI Overview triggers at all, and on which pages Google’s systems select as supporting links. Google tells site owners to apply “the same foundational SEO best practices for AI features as you do for Google Search overall”: technical requirements, policies and helpful, people-first content. Its guide also lists practices such as making sure structured data matches the visible text on the page and that Merchant Center and Business Profile information is up to date.

What this means in practice is that work aimed at AI Overview visibility looks like ordinary Search work:

• answer the question the page targets clearly and early, so a snippet-sized passage exists;
• keep important information in text that Google can index rather than only in images or interactive widgets;
• keep facts consistent across the page, its structured data and its feeds;
• maintain the page so facts that change (prices, dates, product states) are current.

There is no published ranking formula for AI Overview citations, and Google has not documented a separate “AI Overview score”. Claims about specific citation factors or percentages should be treated as a third party’s interpretation unless they cite Google’s own documentation.

Layer 3: control

Site owners have three distinct controls that touch AI features in Google Search. They are not interchangeable.

Snippet rules. Google’s AI features guide says that to limit the information shown from pages in Search, site owners can use nosnippet, data-nosnippet, max-snippet or noindex. These apply across Search, not only to AI features, so a strict snippet limit also affects regular result snippets.

The Search generative AI control. In Search Console, under Settings > Search generative AI, a property can include or exclude its links and content from Search generative AI features. Google’s help page lists AI Overviews, AI Mode and generative AI features in Google Discover, says the control was rolled out to all websites worldwide as of August 31, 2026, and says that it “isn’t used as a ranking or inclusion signal affecting other parts of Search” (Search generative AI control help). Google states that changes generally take a few days to take effect, and that content from other sites remains available in these features and “may appear similar to yours”.

The control has an inheritance model. A property inherits the setting of its closest parent property that has stopped inheriting, and by default the top-level domain property. Before concluding that a subfolder property is included or excluded, open the setting on that property and check whether it says “Inherit control from parent”. Owners of the top-level domain property see a notification when a child property overrides the parent.

Google-Extended. This robots.txt token controls whether content Google crawls may be used for training Gemini models and for grounding in Gemini Apps and Vertex AI. Google’s crawler documentation says it does not affect inclusion or ranking in Google Search (Google’s common crawlers list), and the Search generative AI control page points to it for training decisions. Disallowing Google-Extended is therefore not a way to leave AI Overviews.

Layer 4: measurement

Google gives two views of AI feature performance in Search Console.

The standard Performance report. Google’s AI features guide states that sites appearing in AI features are included in overall search traffic and reported within the “Web” search type. These impressions and clicks are mixed with classic results; they are not separated out in this report.

The Generative AI performance report. Announced on June 3, 2026 and rolled out to all websites worldwide as of August 31, 2026, this report gives “dedicated views of your impressions within generative AI features on Search, such as AI Overviews and AI Mode” (Google’s June 2026 announcement). Its help page documents these dimensions: pages (the final URL linked after redirects, assigned to the canonical), countries, dates in Pacific Time, and devices (Generative AI performance report help).

Read the report with its definitions in mind:

• Impressions only. Google defines an impression here as the number of times links to your site were shown to a user in a generative AI feature. The help page describes impressions; it does not describe clicks, CTR, queries or average position for this report. Do not add those columns to a report and label them as Google AI data.
• Property-level aggregation in the chart. Two links from the same site in one AI response count as a single impression in the chart total; table totals grouped by page can differ.
• Usual Search Console limits apply, including the 1,000-row table limit.
• Search Labs experiments are excluded.
• Discover has its own generative AI report.

Since September 24, 2026, both the standard Search performance report and the Generative AI report can be filtered by multimodal search. Google’s announcement says the data covers searches with Lens, Circle to Search on Android, image uploads to Google Search and Chrome’s “Search this image” feature (Google’s September 2026 announcement). In the Generative AI report the search type filter distinguishes “Web: text-based” from “Web: multimodal”. A site with strong image content can therefore see whether its AI feature impressions come from typed queries or from image-led searches.

A measurement and control matrix

Rising AI impressions, falling clicks: reading the data

Suppose a recipe publisher’s organic clicks fell over two months while impressions in the standard Performance report stayed flat. The owner suspects AI Overviews.

1. Check eligibility. URL Inspection confirms the key recipe pages are indexed; no nosnippet is present.
2. Check the control. The domain property is set to Include; the URL-prefix property for /recipes/ inherits from it.
3. Open the Generative AI report. It shows rising impressions for the recipe pages over the same period, mostly on mobile.
4. Compare with the standard report. Clicks to the same pages fell while impressions held steady.
5. Interpret carefully. The data shows that links to the site appeared more often in AI features while clicks fell. It does not show that AI Overviews caused the click decline, because the AI report has no click data and the standard report does not separate AI clicks. Seasonality, ranking movement and SERP layout changes remain candidate explanations and need their own checks.

The disciplined conclusion is a correlation with a list of alternatives, not a verdict.

AI Overview claims to avoid

• That a specific markup type, file or token makes a page eligible for AI Overviews. Google says none is needed.
• That Google-Extended controls AI Overviews.
• That the Generative AI report shows clicks or queries.
• That a page will be cited if it ranks first. Selection is made by Google’s systems and is not guaranteed.
• That excluding a site from Search generative AI features protects its ideas from appearing; Google’s help page says similar content from other sites may still appear.

The decision in short

Treat AI Overviews as part of Google Search: make pages indexable and snippet-eligible, keep them accurate and useful, decide deliberately whether to stay in or leave Search generative AI features, and measure with the dedicated impressions report while keeping its missing dimensions in view. Training preferences belong to a different control.