Google Search
How to measure Google AI Search visibility in Search Console
A practical guide to Google Search Console's Generative AI performance report: its metrics, dimensions, limits, and a repeatable way to use the data.
Takeaways
- Save a reproducible baseline. Export page and date data with the period, filters and export date. Keep those settings consistent in the next comparison.
- Check the data before explaining a spike. Avoid comparing preliminary data with finalized periods, and check Google's reporting-anomalies page before attributing a change to your work.[1][4]
- Keep impressions, clicks and conversions separate. A displayed link is not a visit. Similar trends across reports do not establish that an AI feature generated a conversion.[1]
- Turn page reviews into specific edits. Find outdated evidence, missing examples or unclear instructions. Record what changed and when; subsequent impression changes do not prove the edit caused them.
- Diagnose a missing report differently from a missing page. Google says low impressions can prevent the report from appearing. For an individual page, check indexing and snippet eligibility before judging its content.[1][3]
What the report measures
Google says the report became available worldwide on August 31, 2026, though a property may not show it if it has too few impressions.[1]
The report includes AI Overviews and AI Mode. Its dimensions include:
- Pages: the canonical page URL linked from an AI feature.
- Countries: where the search originated.
- Dates: daily, weekly, or monthly groupings.
- Devices: desktop, tablet, or mobile.
The Search report covers text-based and multimodal Web searches. Search Console excludes data from active Search Labs experiments. The newest chart data can be preliminary and may change as Google finishes collecting it.[1]
Google defines these impressions as appearances of links to a site in a generative AI feature. Chart totals and table totals can differ because the chart and table may aggregate at different levels, such as the property or page.[1]
What the report cannot tell you by itself
The report is organized around impressions and the dimensions listed above. It does not provide a complete transcript of the prompts or query paths behind each appearance. That means it can identify pages with exposure without explaining the exact user question that led to it. This is an inference from the report’s documented fields, not a claim about Google’s internal data.[1]
An impression is evidence that a link appeared. It does not measure whether a reader clicked, understood the answer, trusted the brand, or became a customer. Use broader Search Console performance data and analytics to study traffic and outcomes. Keep those measures separate from the dedicated AI impression count.[1][2]
A repeatable review
1. Confirm the report is available
Open the Search Console property for the canonical site. If the report is missing, Google says the property may not have received enough impressions in generative Search features.[1]
2. Record a baseline
Export the data by page and date. Save the export date, reporting period and country or device filters. Compare periods of the same length with the same filters, and note publishing dates or seasonal events that could affect the comparison.
3. Review pages with the most impressions
Sort by impressions and read the pages at the top of the list. Check whether their answers, examples and evidence are current and complete. Record the gaps you can fix without treating any passage as the confirmed reason Google selected the page.
4. Compare visibility with visits
Compare broader Search Console performance and site analytics for the same pages and period. Report AI impressions, Search clicks, sessions and conversions separately. Parallel trends do not establish which visits or conversions came from an AI feature.
5. Fix documented reader problems
If a page is getting impressions but serving readers poorly, improve the evidence, examples, or next steps. If a useful page has no visibility, check indexing, snippet eligibility and the effective Search generative AI inclusion setting using the site owner's guide. Google does not guarantee indexing or serving, even when technical requirements are met.[3]
Reading changes carefully
Check Google’s Search Console data anomalies page before explaining a sudden dip. For example, Google has documented resolved logging errors affecting AI performance report impressions in 2026.[4]
What to report to a team
Use a compact dashboard with:
- AI-feature impressions by page and period.
- Standard Search clicks and impressions for the same pages.
- Sessions and conversions by recorded source/medium for the same pages and period.
- Editorial changes and technical incidents recorded on a timeline.
- A note that exposure does not establish causation or future inclusion.
Include the reporting period, filters, page changes and unresolved data issues. Do not attribute conversions to AI features from impression trends alone. For eligibility checks, use the site owner’s guide.
Choose the questions and page requirements before interpreting changes in impressions. The AI search prompt strategy guide help build an initial prompt universe and distinguish retrieval, citations, recommendations and customer outcomes.
Report identifiable assistant visits separately
For assistant referrals beyond Google Search, use the GA4 attribution workflow. It walks through the AI Assistant channel, entry pages, a selected registration key event and downstream matching. Keep that report separate from Search Console and Google AI-feature measurements.
Sources
- Generative AI performance report (Search). Google Search Console Help.
- Introducing Search Generative AI performance reports in Search Console. Google Search Central Blog.
- Google Search technical requirements. Google Search Central.
- Data anomalies in Search Console. Google Search Console Help.
Last verified October 17, 2026.