How can a local business measure [AI Overview visibility](/answers/choosing-a-local-rank-tracker-for-ai-search) and clicks?

What does Search Console show for AI Overview traffic?

The existing answer “AI Overviews in Local Search and Clicks” keeps the bounded appearance studies and click-loss evidence. This answer is the canonical measurement procedure and supersedes that page's shorter Search Console section and date-matched check. For a separate Bing view, see the Topics Portfolio beta, which reports approximate citation themes rather than Google Search Console measurements.

Google's Generative AI performance report (Search) reports impressions for AI Overviews and AI Mode. Google says the report rolled out to all websites worldwide by August 31, 2026, and lets you group those impressions by pages, countries, devices, and dates. Search Labs experiments are excluded.

The Generative AI report counts impressions when links to your site are shown in a Google Search generative AI feature. It does not count every time a business is named or cited without a link. Google assigns most data to the page's canonical URL, but its Search Console metric guidance says that some data can be assigned to the actual URL instead.

Use the Performance report (Search results): Overview and basic setup for Web-report clicks, impressions, click-through rate, and average position. A click on an external page from an AI Overview or AI Mode counts as a click, and Google includes the Generative AI report's data in the Web search type. The two reports answer different questions: AI-feature impressions show linked visibility in generative features, while Web clicks show visits from Search without proving which surface produced each click.

For the definitions behind the report, see the guide to Google Business Profile performance metrics. For a broader testing routine across Google and ChatGPT, see how a business can appear in AI search results.

What should I record for each AI Overview measurement?

Keep one row for each direct observation and one matched set of Search Console exports. The following fields let you connect an appearance check to a click outcome without treating them as the same event.

An observation of one query is evidence about that query, location, device, surface, and date. Repeat the check with the same context and keep the source links when they are shown. If the result contains local business data, where ChatGPT gets local business data explains why a provider's cited sources and a Google Business Profile should be checked separately.

How do I run a date-matched measurement check?

Use the same cohorts and date windows throughout the check:

1. Freeze the query and page cohorts. Keep the exact queries and landing pages. Split the sample by intent when it is large enough to support that distinction. Record the business location and device used for direct checks.
2. Run direct appearance checks. For every fixed query, record the timestamp, location, device, whether an AI Overview appeared, whether the business or page was named or linked, and the cited URL when visible. Repeat the checks because local results can vary with context.
3. Export both first-party reports. For the same date windows, export AI-feature impressions from the Generative AI performance report by pages, countries, devices, and dates. Export Web-report clicks, impressions, click-through rate, and average position for the query and page cohorts. Keep the date range with both exports. Google's Performance report guidance says the newest data can be preliminary and may change during the following hours.
4. Mark dated context. Record a Google rollout, site change, seasonal shift, analytics change, or other plausible event that overlaps the window. Google's May 1, 2025 AI Mode product announcement is historical rollout context, not a statement of current availability.
5. Choose a non-AI comparator. Use a stable query or page group exposed to the same site, season, and reporting changes but not treated as AI-affected. Compare the groups over matched dates, not one group before a change and the other after it.
6. Write the attribution at the strength the log supports. If the observation window, query and page groups, comparator, and competing explanations line up, report an association in time. If they do not line up, report the Web movement and the AI observation as separate facts.

What do local AI visibility studies actually measure?

Third-party studies can show how often a feature appeared in a defined sample. They are useful context for designing repeated, localized checks, but their rates are not a current prevalence figure for every business or market and they do not replace your Search Console data.

Use the study date, population, query design, location method, and surface alongside any reported percentage. The study's number describes its collection method, not what a local business should expect from its own queries today.

How should I describe a change in clicks without claiming AI caused it?

Describe the recorded observations first, then state the relationship that the matched comparison supports.

The strongest defensible conclusion is an association between a dated appearance observation and a matched Web movement, after competing explanations have been checked. A universal click-loss percentage needs stronger evidence than the appearance studies listed here.