Google AI Overviews: what you can measure and what you cannot

What AI Overviews and AI Mode are

AI Overviews are the AI-generated summary Google shows above some search results. AI Mode is Google Search's conversational experience, built for questions that need exploration, reasoning or comparisons. According to Google's page AI features and your website, both surface relevant links to help people find information and explore content they may not have discovered before.

Three points from Google matter for measurement. First, both features may use a “query fan-out” technique: they issue several related searches, across subtopics and data sources, to build the response. Second, AI Mode and AI Overviews may use different models and techniques, so the responses and links they show vary. Third, an AI Overview is only shown when Google's systems determine it adds something to classic Search, so it often does not trigger.

The direct consequence: there is no single “position” in AI Overviews the way there is a position in the blue links. The same query may show an overview or not, and cite one page or another, depending on the moment and on the related searches it ran.

Eligibility: what Google actually requires

The rule is simpler than a lot of advice suggests. To be shown 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, meeting Search's technical requirements. Google states there are no additional technical requirements, and that meeting every requirement does not guarantee crawling, indexing or serving.

• No special markup: Google says there is no special schema.org structured data you need to add for these features.
• No dedicated file: its guide to optimizing for generative AI features lists LLMS.txt and other “AI text files” among the things you can ignore for Google Search, which does not use them.
• No artificial chunking of content and no rewriting “for AI”: Google says its systems understand several topics on one page.
• One more condition: according to the same guide, in addition to Search's technical requirements, a site must be included in Search generative AI features in Search Console to be eligible for display.

In other words, the levers are still SEO levers: a page that is crawlable, indexable and useful, with its important content available as text. To check those basics on your site, our article on the technical SEO audit walks through the checks that matter.

What you can measure

Search Console's Performance report

Google puts it plainly: like the rest of the results page, sites appearing in AI features (AI Overviews and AI Mode) are included in the overall search traffic in Search Console, specifically in the Performance report, within the “Web” search type. Your clicks and impressions from AI Overviews are therefore already in your figures: nothing is lost, but nothing is isolated either.

The Generative AI performance report

Google's guide recommends the Generative AI performance report in Search Console to see how your content performs in generative AI features on Google Search and Discover. It is the only dedicated view Google offers. We do not describe its metrics here: read them in your own account, since they can change.

What happens after the click

Google also suggests tracking conversions and time spent on your site in other tools, such as Google Analytics. That is where you judge the real value of a visitor who came from search, whether or not they went through an AI Overview.

What you cannot measure

• Isolating AI Overviews inside “Web”: the Performance report counts them together with classic results. A rise or drop in clicks does not, on its own, tell you whether an AI Overview caused it.
• Querying AI Overviews programmatically: Google offers no API for its AI Overviews. A tool that claims to “track” them automatically should explain where its data comes from.
• Drawing a conclusion from one check: answers vary with the models, the related searches and whether the overview triggers at all. A screenshot is a dated anecdote, not a trend.
• Reaching Google's internal signals: Google warns about third-party tools that promise ranking success or claim to use “internal” metrics; no third-party tool has access to its ranking or AI systems.

The controls you have

Measuring also means knowing what you can decide. According to Google, AI is built into Search: robots.txt directives for Googlebot remain the way to manage how your site is crawled for Search (see the introduction to robots.txt).

• To limit what is shown from your pages in Search: nosnippet, data-nosnippet, max-snippet (see controlling snippets) or the noindex rule.
• To limit AI training and grounding in some of Google's other systems, Google points to Google-Extended, described among its common crawlers. It is not a visibility setting for Google Search.

An honest measurement protocol

1. Pick a short, stable list of queries that matter to your business: category questions, comparisons, problems you solve.
2. For each one, record on a fixed date: does an AI Overview show, is your site cited, and with which URL.
3. Read those checks alongside the “Web” trend in Search Console and your conversions, without assigning a change to a single cause.
4. Check again after each significant content change, and compare across several dates before concluding.

How NeoRank handles AI Overviews

NeoRank does not query Google and does not pretend to. The “AI Overviews (Google)” tile follows exactly the two legitimate paths described above, as the documentation on the engines queried explains.

• Search Console: connected to your own account (see integrations), it already includes AI Overviews in the “Web” type, with no way to isolate them.
• Manual reading: “Check in Google” opens the tracked prompt in your browser. You then record whether an AI Overview shows, whether your site is cited, the cited URL and the date.
• The tile then shows the share of checked queries with an AI Overview and the share citing your site, labelled “manual reading — not automated”, dated, and never mixed with the AI engines' observations.
• A reading older than 30 days is flagged stale; with no reading, the tile stays “Not measured”, never zero (see what “Not measured” means). A client viewer cannot record a reading.

The per-engine tiles are described in AI visibility. For the engines NeoRank really queries, such as ChatGPT or Perplexity, read our guide to getting cited by answer engines.