← The field guide

Part of the AI search measurement collection ↗

How to read Search Console’s generative AI report

Measure Google AI Overviews and AI Mode exposure, avoid double counting, and turn an impression report into a useful content decision. Includes a review worksheet.

The short answer

Use Search Console’s generative AI report to measure exposure in Google’s AI search features. Keep that observation beside overall search clicks and collected site visits. An AI impression establishes visibility; it does not establish a visit, a lead, or an attributable sale.

What the new report tells you

Google announced dedicated generative AI performance reports on June 3, 2026, and marked the worldwide rollout complete on August 31. The Search view covers AI Overviews and AI Mode; Discover has a separate report. Search Labs experiments are excluded.

The documented metric is impressions, with page, country, device, and date breakdowns. This gives an editor a direct observation to act on: a particular resource is appearing in AI search. It does not supply a separate AI-click numerator for calculating AI CTR.

Source: Google Search Central: Generative AI performance reports; Search Console: Generative AI performance report (Search)

Run a repeatable page review

Open the correct Search Console property and the generative AI performance report for Search. Use the same completed date range on every saved view. For a small publication, two adjacent 28-day windows make a more useful editorial comparison than yesterday versus today.

Record the chart total first, then inspect Pages. Select a relevant canonical URL and save its filtered report separately. Inspect countries and devices only when those cuts would change the audience or implementation decision.

  • Keep the report name, property, dates, timezone, filters, aggregation, and export time with the numbers.
  • Check whether the latest point is preliminary; revisit it before treating a small change as a trend.
  • Preserve the page report and the query evidence from ordinary Search separately. Do not assign a query to an AI appearance without evidence.

Source: Search Console: Dimensions and data groupings; Search Console: About the data

Do not add AI impressions to Web impressions

Google includes this visibility in overall Search performance. If a fictional site records 2,000 Web impressions and 120 generative AI impressions, its Web total remains 2,000. Reporting 2,120 would count overlapping exposure twice.

Aggregation also matters. Imagine two different articles from one site appearing together in one result set. A property total can count one impression while two page rows each receive one. Keep page-level and property-level totals labeled; adding rows is not a repair for a mismatched chart.

Source: Google Search Central: Generative AI performance reports; Search Console: About the data

An example that leads to an editorial decision

These invented rows cover September 1–14, 2026, for example.com. The editor compares each page only with itself under matching filters. The first article has the strongest evidence of discovery, so its next improvement is a worked answer to the reader’s task, supported by the query review.

The 14 clicks are overall Web clicks. Their overlap with the 80 AI impressions is unknown. Dividing 14 by 80 would not produce an AI click-through rate.

Fictional pageWeb impressions / clicksAI impressionsNext action
/guides/measurement800 / 1480Add a concrete interpretation example; preserve its existing URL
/guides/feed-errors300 / 025Review page-filtered queries before rewriting the title
/guides/new-checklistNot yet availableNot yet availableCheck report coverage and index status; retain unknown values
Download the fictional page-review worksheet (CSV)

A missing report is not a measured zero

If the report or value is unavailable, keep the value unknown and record the limitation. Google notes access, impression volume, and eligibility as possible reasons a report is absent. A blank interface is insufficient evidence that no AI system has ever used your page.

Exports can turn unavailable values into zero. The worksheet therefore stores both a value and a status. Retain a screenshot or the original report alongside the export so that a later comparison cannot silently convert “unavailable” into “observed zero”.

Download the examples, blank record, and review rules (JSON)

Source: Search Console: Generative AI performance report (Search); Search Console: About the data

Choose the next improvement from the evidence

When a useful page gains exposure, improve the answer that makes it worth opening: a calculator, a source-linked comparison, a reproducible check, or a filled template. Keep a dated change log and compare subsequent completed periods. This is an editorial test; one before-and-after observation cannot isolate the effect of the edit.

Use overall Web clicks to judge search visits and your collected site data to examine behavior. Keep the AI report as its own exposure measure. Expand a topic when readers show interest or a specific unanswered task is visible; page volume alone is not the objective.

Sources & scope

Primary references checked for this edition. The notes below distinguish source-backed facts from the frameworks and examples proposed in this guide.

  1. Google Search Central: Generative AI performance reports ↗

    Confirms the August 31 worldwide rollout, available dimensions, and inclusion within overall performance. It does not provide a traffic forecast.

    Checked September 18, 2026
  2. Search Console: Generative AI performance report (Search) ↗

    Defines the Search report, impression counting, availability, and export limitations. Checked against the current documentation.

    Checked September 18, 2026
  3. Search Console: About the data ↗

    Explains property versus page aggregation, preliminary data, and unavailable values in exports.

    Checked September 18, 2026
  4. Search Console: Dimensions and data groupings ↗

    Supports canonical-page reporting, omitted queries, filters, and reporting time boundaries.

    Checked September 18, 2026

AI-assisted research and drafting. Provider-specific claims link to primary sources. Frameworks are editorial proposals; worked examples are illustrative and are not employer performance results.

Editorial policy & corrections ↗

Cite this guide

George Kelly. How to read Search Console’s generative AI report iamgeorgekelly. Updated September 22, 2026. https://www.iamgeorgekelly.com/field-guide/search-console-generative-ai-report

Keep the source notes and example labels with an excerpt. For a provider requirement, follow the original documentation in Sources & scope.

Choose your next task ↗ · Browse the source directory ↗

Keep going

How to measure AI citations and visible brand attributionHow to find referral traffic in GA4, including AIAI visibility metrics: impressions, citations, referrals, and outcomesSearch impressions but no clicks: a diagnostic and planning calculatorWhy Search Console, GA4, and Vercel show different traffic numbers