Google Search Console now has dedicated generative AI performance reports for selected websites. The new reports show impressions from generative AI features in Search and Discover, including which pages appeared and how visibility changed by country, device, and date.
Google announced the rollout on June 3, 2026. It is a meaningful step because site owners can finally separate part of their Google generative visibility from the broader Performance report.
It is not a complete GEO dashboard. It does not tell you which prompts produced the impression, what the generated answer said, whether your brand was recommended, or which competitors appeared. The report is valuable precisely when teams understand both its strengths and its boundaries.
What Google announced
Google says it is rolling the reports out to a subset of websites for testing and feedback before wider availability. There are dedicated views for generative AI features in Search and in Discover.
For Search, the included capabilities currently cover AI Overviews and AI Mode. Google says the data remains part of the overall Performance report as well; the new experience provides a separate view of that visibility.
The report contains the following dimensions and information:
- Impressions: how often URLs from the property appeared in supported generative AI features;
- Pages: which canonical URLs appeared;
- Countries: where the visibility originated;
- Devices: desktop, tablet, or mobile for Search;
- Dates: trends over supported hourly, daily, weekly, and monthly granularity.
Google’s help documentation says the Search report currently contains impression data, not a complete set of traditional performance metrics. Teams should read the current interface and documentation rather than assuming that clicks, click-through rate, or query data are available in the same way.
How impressions are counted
The details matter when numbers become targets.
Google says the chart aggregates generative AI impressions by property. If two results from the same site appear in one generative feature, the property-level chart can count that as one impression. When a URL filter is applied, aggregation changes to the URL level.
The table changes aggregation by dimension. Country, device, and date use property aggregation, while page data uses page aggregation. Because of that, chart totals and table totals can differ.
Most page data is assigned to the canonical URL, not a duplicate URL. The newest data can be preliminary, and standard Search Console limits—including the 1,000-row table limit—apply.
None of these details makes the report unreliable. They mean analysts should document the view and aggregation before comparing totals.
Why this is important
Until now, teams often had to infer Google AI visibility from mixed Search performance data, referrals, manual observations, or third-party monitoring. The dedicated report provides first-party evidence that a site’s URLs appeared within supported Google generative features.
That creates several useful workflows:
- identify the pages receiving the most generative AI impressions;
- compare countries and devices;
- monitor a stable trend over time;
- connect visibility changes with page updates or editorial work, cautiously;
- find pages worth inspecting in AI Overviews and AI Mode.
It also establishes a healthier measurement boundary. Google owns the impression count for its product. Third-party platforms can complement that data with prompt and answer observations, but should not present their estimates as internal Google metrics.
What the report does not answer
It does not expose the prompt or query
Based on the announced dimensions, teams cannot see which user question caused a page to appear. That limits intent analysis and makes it difficult to connect an impression to a specific buyer journey stage.
It does not show the generated wording
An impression says a link appeared. It does not tell you how the answer framed the topic, whether the brand name was present, or whether the description was accurate.
It does not provide competitor context
Search Console reports on your verified property. It does not tell you which competitors or third-party sources appeared beside you.
It does not prove causation
If impressions rise after a page update, that is a useful observation, not proof that one edit caused the change. Search demand, feature availability, crawling, indexing, competitors, and Google’s systems may all change.
It does not cover other AI platforms
The report is for Google Search and Discover. It says nothing about ChatGPT, Gemini Apps outside the reported Search surfaces, Claude, or Perplexity.
How to use the new report
1. Confirm scope and access
Look for the generative AI report in Search Console. If it is missing, Google says the property may not be included in the limited rollout or may not have enough impressions. Record the date access begins.
2. Establish a clean baseline
Export the available data and preserve the date range, filters, aggregation, and time zone. Search Console dates use Pacific Time. Annotate site migrations, major publishing changes, and known reporting issues.
3. Find high-impression pages
Review the top pages and classify their role: guide, product page, comparison, documentation, news, or other. Look for shared traits without assuming they are a formula.
4. Find important pages with low visibility
Compare the report with strategic pages in your sitemap and traditional Search performance. A low number can reflect limited demand or feature triggering, not necessarily poor content, so investigate before rewriting.
5. Combine first-party and observed answer data
Use Search Console for Google impression trends. Use a consistent prompt library to inspect answer wording, brands, competitors, and citations across relevant platforms. Our AI visibility measurement guide explains how to preserve that context.
6. Connect visibility to outcomes
Where referral and analytics data are available, inspect engaged sessions, sign-ups, assisted conversions, or other meaningful outcomes. Google has said that clicks from AI Overviews are included in Search Console’s overall Web performance data, but the dedicated report’s current focus is impressions. Avoid inventing a click rate from incompatible totals.
A useful combined dashboard
A practical weekly or monthly view can include:
- Google generative AI impressions and trend from Search Console;
- top and emerging pages by Google AI impressions;
- brand mention and recommendation rates from the stable prompt sample;
- owned and third-party citation domains observed in answers;
- competitor presence by buyer journey stage;
- material accuracy issues;
- organic visits and conversions where attribution is available;
- the next three evidence or content actions.
Keep each metric’s source visible. First-party impressions, sampled answer observations, and web analytics describe different parts of the journey.
Where Swep fits
Search Console should be the source of truth for the Google impression data it exposes. Swep adds prompt-level monitoring and competitive context across supported AI surfaces: which answers mention the brand, what competitors appear, what sources are cited, and where recurring gaps suggest work.
That combination is more useful than either view alone. Search Console can show that a page appeared more often. Swep can help a team inspect the answer landscape around the questions it deliberately monitors. Neither can guarantee a future inclusion or isolate causation without careful analysis.
Use the documentation to align your prompt groups with product lines and markets, and review the guide to AI visibility metrics before combining signals.
The takeaway
Google’s June 3 announcement gives website owners a valuable first-party view of generative AI impressions. Use it to understand trends and pages inside Google’s supported features.
Do not ask it to answer questions it was not designed to answer. Pair it with prompt-level observations, competitor analysis, source review, and business outcomes. The dedicated report is not the end of AI visibility measurement. It is a much better foundation for the Google part of it.