Search Console Generative AI Report: How to Measure AI Overviews and AI Mode

Google Search Console can now provide a separate view of impressions earned by your URLs inside generative AI features. For the first time, a team can open an official Google signal, identify which pages appeared, and segment the data by country, device and date without inferring it from a third-party tracker.

The change matters, but the number is easy to overstate. An impression is not a visit, recommendation or conversion. The report does not reveal the query, answer, cited passage, or whether an appearance occurred in AI Overviews or AI Mode. It measures one specific layer: URL exposure inside Google's generative Search experiences.

This guide explains what the beta report contains, how its aggregation works, which decisions it supports, and how to combine it with independent measurement. If you first need a manual result-checking method, use the guide to measuring brand mentions in AI Overviews. Here, the unit is different: Search Console's official aggregate report.

What Changed in June 2026

Google announced its generative AI performance reports on June 3, 2026. The company introduced dedicated Search and Discover views to isolate impressions generated when site URLs appear in AI experiences.

The Search view covers features such as AI Overviews and AI Mode. Google says this data is already included in the overall performance report; the dedicated view makes it separately analyzable. The product remains in beta and is rolling out to a subset of sites, so two verified properties may not have identical options.

This moves the conversation from "we think Google uses us" to "Google counted these appearances for these URLs and conditions." It does not complete attribution. It adds first-party platform evidence that must be stored with its definition and limitations.

What the Search View Includes

The report observed for Mentio provides one primary metric and four operational dimensions:

This is not an answer inventory. It is an aggregate performance table. To see whether your brand appears, which competitors surround it and how it is framed, you need additional observations such as an AI visibility tracker.

What an Impression Means and Does Not Mean

Google defines an impression as an occasion when a site URL appeared in a generative AI feature. Preserve that wording in your metric dictionary. Do not rename it "mention," "citation" or "recommendation," because those events need different evidence.

An impression supports these statements:

• a URL was eligible and appeared under Google's measurement;
• exposure occurred inside the selected range and filters;
• it can be assigned to a page and analyzed through available dimensions.

It does not support these statements:

• the person clicked or visited the site;
• the URL was the primary link or visible without interaction;
• Google named the brand in generated text;
• the answer was positive, accurate or commercial;
• the impression came specifically from AI Overviews or AI Mode;
• the appearance produced a conversion.

Use explicit field names such as gsc_genai_impressions, not ai_mentions. This keeps a dashboard from turning potential reach into a business outcome.

AI Overviews and AI Mode Are Aggregated

The Search report includes both experiences, but the beta version we observed has no surface dimension. If one URL records 100 impressions, you cannot split them between AI Overviews and AI Mode. Do not estimate the split from position, device or ordinary Search queries.

Google's official guide explains that generative Search experiences use core Search systems, page retrieval and query fan-out. The report does not expose the original query or related system-generated queries. A page may appear for a semantically connected need without the table revealing the path.

When a decision requires surface-level comparison, run a controlled sample and inspect each result independently. The guide to brand visibility in Google AI Overviews covers that layer; do not replace it with an inference from Search Console.

How Aggregation and Filters Work

The report documentation distinguishes aggregation levels. The total chart is property-aggregated. The pages table is URL-aggregated, and a URL filter shifts analysis to that page. Country, device and date describe the property under active filters.

As a result, the chart total may not equal the sum of visible table rows. Do not repair that difference by adding rows or report it as missing data. Record:

• property;
• report type: Search or Discover;
• range and time zone;
• active filters;
• open dimension;
• export date and time;
• whether the newest data was marked partial.

Use dated exports. A "last 28 days" row changes daily, and recent data may be revised. Compare equivalent closed windows or preserve every snapshot so the value seen by the team can be reconstructed.

What Is Missing and Why It Matters

For Mentio, the beta view does not expose queries, clicks, CTR or position. It also omits generated text, selected passage, visual order, individual surface and present competitors.

Those omissions define the permitted questions:

• Yes: Which pages concentrate generative Google impressions?
• Yes: Is exposure increasing in one country or device?
• Yes: Has a new URL entered the visible page set?
• No: Which exact query triggered the appearance?
• No: What percentage of impressions generated a click?
• No: Did Google recommend our brand over a competitor?
• No: Was a decline caused by AI Overviews or AI Mode?

The ordinary Web performance report retains its own query, click, CTR and position data, but do not join those fields to a generative impression as if a one-to-one attribution key existed. Compare trends and pages cautiously; do not manufacture a relationship the product does not provide.

Real Example: 8,296 Impressions for Mentio

On August 16, 2026, Mentio reviewed its domain property for July 18 through August 14. The report showed 8,296 generative AI impressions across 54 pages.

Exposure was concentrated:

The first URL represents approximately 79.7% of the total. The valid conclusion is: "observed generative exposure is highly concentrated on one page during this window." It is not valid to claim that the page generated 79.7% of AI visits, recommendations or conversions.

The next action is not an immediate rewrite of the winning page. First check whether concentration persists, whether it depends on country or device, whether other URLs are growing, and whether the pattern agrees with direct answer observations. The number opens an investigation; it does not provide the cause.

A Weekly Workflow for Turning the Report Into Decisions

1. Freeze the Reading

Export on the same weekday and record range, filters and update status. Preserve every file. If you use a rolling 28-day window, add a comparison with the previous 28 days.

2. Normalize Pages

Resolve protocol, host, parameters and redirects, then assign each row to its canonical. Retain the original value for audit. Group by page type only afterward: article, industry, product, home or another template.

3. Calculate Concentration and Coverage

Track total impressions, active pages, top-five share, median per page, and pages entering or leaving the set. A total increase carried by one URL creates a different risk from broad corpus growth.

4. Segment Before Explaining

Open country, device and date. Identify localized movement. Do not confuse country with language or device with answer format. Annotate releases, migrations, publications and canonical changes on the series.

5. Contrast Other Layers

Join by page and period, never by a fictional individual impression. Review indexing, organic traffic, conversions, logs and controlled prompt observations. When impressions rise but brand mentions do not, Google may be using the URL as a source without naming the brand, or exposure may occur under another intent.

6. Write a Testable Hypothesis

Example: "the new guide expanded generative exposure on Spanish mobile after indexing." Define what would support or reject the hypothesis in the next window. Avoid acting on a single screenshot.

Decision Matrix

Do not turn every pattern into a content task. Sometimes the correct action is to wait for another window, repair a canonical, expand measurement, or investigate a Google product change.

Combining Search Console With an Independent Tracker

Search Console and a tracker do not share one definition:

• Search Console: official Google source, aggregated URL exposure, real coverage not limited to a prompt bank.
• Tracker: designed sample, observable answers, prompts, mentions, rank, competitors, sentiment and multiple assistants.

Keep them as separate columns. Never add GSC impressions to tracker runs or calculate a shared percentage. Look for convergence: a page with rising exposure and more mentions inside a stable sample provides stronger evidence than either signal alone.

Manual AI Overview mention measurement helps inspect answers; the AI visibility tracker adds repetition and comparison. Search Console adds Google's view of exposed URLs. The three layers use different units.

Common Mistakes

• Calling impressions "AI traffic."
• Forcing table rows to equal the chart without checking aggregation.
• Attributing a change to AI Overviews when AI Mode is included.
• Inferring generative queries from the ordinary Web report.
• Comparing rolling ranges captured on different dates.
• Duplicating parameterized URLs or ignoring canonicals.
• Optimizing only the leading page and increasing concentration.
• Treating a missing report as proof of no visibility.
• Using partial data as an alert without rechecking update status and filters.
• Replacing answer observation with an aggregate metric.

Implementation Checklist

• [ ] The correct property is verified and Search is separate from Discover.
• [ ] Every export stores range, filters, time zone and update status.
• [ ] "Generative impression" keeps Google's definition.
• [ ] URLs map to canonical while retaining the original value.
• [ ] Total, active pages and concentration are reported.
• [ ] Country, device and date are reviewed before assigning causes.
• [ ] No query, click, CTR, position or surface is invented.
• [ ] Partial data is labeled and checked again.
• [ ] GSC, tracker and business outcomes remain separate layers.
• [ ] Every action starts from a hypothesis and validation window.

FAQ

What does the Search Console generative AI report measure?

It measures how often property URLs appear in generative AI features on Google Search and lets you analyze those impressions by page, country, device and date. It is an exposure signal within Google, not a direct measure of clicks, recommendations, sentiment or conversions.

Does the report include AI Overviews and AI Mode?

Yes. Google says the Search view includes features such as AI Overviews and AI Mode. In the observed beta version, they are aggregated: the report does not provide a breakdown that assigns an impression to one specific surface.

What does a generative AI impression mean?

It means a site URL appeared in a generative AI feature under Google's measurement rules. It does not prove that the person visited the page, read the link, received a favorable recommendation, saw a prominent citation or completed a conversion.

Why can I not see queries, clicks, CTR or position?

The beta report we reviewed shows impressions plus page, country, device and date dimensions, but not queries, clicks, CTR or position. Those omissions prevent exact intent analysis or traffic calculation from this view. Google has said it is considering additional metrics over time.

Why is the report missing from my property?

Google is rolling out generative AI reports to a subset of websites during testing. A missing view does not prove that a property never appears in AI Overviews or AI Mode; it may simply mean that the property is not yet part of the rollout.

Does this report replace an AI visibility tracker?

No. Search Console provides an official, aggregated signal about URL exposure in Google's AI features. An independent tracker can observe prompts, mentions, position, competitors, sentiment and other assistants. They answer different questions and should be reconciled, not mixed as if they measured the same event.

Turn Exposure Into a Useful Series

Start with a dated export, normalize pages and calculate concentration. Then segment, compare with observed answers and write a hypothesis. The report becomes useful when you can explain what it measures, what it omits and what additional evidence is required before action.

Measure your brand's AI visibility with Mentio ->

Want to know if AI mentions your brand?

Discover your visibility in ChatGPT, Claude and Gemini in minutes.

Analyze your brand for free

Related articles