On June 3, 2026, Google announced new Search Generative AI performance reports in Search Console, with dedicated views for AI Overviews, AI Mode and Discover’s AI features. The reports give a clearer baseline of how URLs surface inside AI-driven features — but they don’t prove traffic loss, ranking change, or AI citation quality on their own.
What the new Search Console report actually is
The announcement is short but specific. Google says it is launching dedicated reports for visibility inside generative AI features on Search — AI Overviews and AI Mode — and a parallel report for AI features in Discover. The data also continues to live inside the overall performance report, so the new view is an additional, dedicated lens rather than a fully separate dataset.
A few details worth memorising before you open the report:
- Rollout is partial. Google says it is releasing these reports “to a subset of websites” first while it tests and gathers feedback. Don’t assume your account already has them.
- Surfaces are explicit. The Search report covers AI Overviews and AI Mode. The Discover report covers generative AI features inside Discover.
- Dimensions are familiar. Impressions, pages, countries, devices (Search only), and dates with hourly, daily, weekly and monthly granularity — the same shape as the existing performance report.
None of that is the same as “Search Console now fully separates AI Overview clicks from ordinary clicks.” That phrasing is being thrown around online and the announcement does not say it. The honest summary: there is now a dedicated view, and the data still rolls up into the overall performance report.
What Search Console can prove — and what it cannot
The new report inherits the same general truth that has always applied to Search Console: it is a Google-side measurement of how URLs appeared in Google surfaces. It is the most authoritative source for impressions and clicks coming from Google. It is not, and has never been, a complete model of cause and effect.
What it can tell you
- How often URLs from your site appeared in generative AI features on Search and Discover (impressions).
- Which specific URLs were shown (pages).
- Where, by country, those impressions happened.
- Which devices people were on (Search-side data only).
- How those numbers move across time, at granular intervals.
What it cannot prove on its own
- Why a particular URL was — or was not — selected by a generative feature.
- Whether a movement in impressions was caused by content change, ranking change, AI feature change, or query-mix change.
- How any AI surface paraphrased, summarised or cited your content downstream.
- Anything about downstream clicks or user behaviour after the impression itself.
This is why the new report is best treated as a higher-resolution input, not an answer machine. The interpretation work stays your job.
Why AI Overview measurement numbers need careful interpretation
A separate, frequently-cited reference point for AI Overviews comes from academic research rather than from Google itself. In a 2026 measurement study, Xu et al. issued 55,393 trending queries to Google Search and decomposed the AI Overview responses into 98,020 atomic claims. In that measured query set, AI Overview activation was 13.7%, and 11.0% of atomic claims were unsupported by the cited pages.
Two cautions about quoting any of those numbers:
- They describe the study’s query sample, not all of Google Search. The query mix in the study is trending queries; your topic mix is almost certainly different.
- They describe what was true when the preprint was written. AI Overviews quality and triggering behaviour change over time.
The same logic applies in reverse to your own Search Console export. A sharp drop in impressions across one month for one URL cluster is a real number in a specific window — not a Google-wide trend, and not on its own evidence that the cause was AI Overviews.
A safe walkthrough for reading the report
Whether your account already has the new generative AI views or not, the workflow looks the same: stabilise the baseline, isolate the variable, then commit to an evidence note before you act.
If your account is in the rollout
- Open Search Console → Performance and look for the new generative AI performance reports beside the existing Search and Discover entries.
- Export a baseline window for both the new AI view and the overall performance report at the same date range. The two are not duplicates — they answer different questions.
- Pivot by page first, then by query. The page view tends to expose which clusters are genuinely surfacing inside AI.
- Annotate every significant change in your tracking sheet: content updates, structured-data changes, internal-link changes, Google rollout dates.
If your account is not yet in the rollout
- Use the existing Performance report and segment by query to surface candidates that look like AI Overview triggers (longer, question-shaped, comparative).
- Manually inspect those queries — incognito, signed-out, and from a representative country and device.
- Record whether your URL appears inside AI Overviews or AI Mode for that query, and which URLs do.
- Re-check on a fixed schedule (for example, weekly), so observed change is not just intraday volatility.
Reading the report at a glance
The table below summarises what each signal you can pull from Search Console — old or new — is actually telling you, and what it is not.
| Signal / Metric | What it can tell you | What it cannot prove | Safe action |
|---|---|---|---|
| AI impressions per page (new report) | How often the URL surfaced inside AI features | Whether it was paraphrased, cited or clicked | Use as a relative baseline; compare to your own prior weeks, not to other sites |
| AI impressions by country | Where your AI visibility is concentrated | Causality of the geographic distribution | Prioritise localisation work where AI visibility is already non-zero |
| Devices (Search side) | The device mix Google saw for your URLs | Why mobile vs desktop differs | Cross-check with GA4 device data to spot tracking gaps |
| Dates (hourly / daily) | Short-window pattern changes | That a change is permanent or caused by your edit | Wait for at least one full weekly cycle before reacting |
| Overall Performance report | Total Google-side impressions and clicks | How AI vs non-AI surfaces split downstream | Treat as the umbrella; use the new view to look inside it, not instead of it |
A practical AI visibility checklist
The checklist I run with clients before changing any content based on AI visibility data:
- Export a baseline from Search Console (Search + Discover + the new AI view if available) before changing anything.
- Compare only stable time windows — at minimum week-over-week, with the same day count.
- Annotate every content, technical and structured-data change with date and scope.
- Separate ranking movement from AI citation movement; they are different problems.
- Validate AI surface presence manually for high-value queries, from a clean session.
- For deeper analysis, export GA4 raw events to BigQuery — Google’s GA4 BigQuery export gives SQL-grade visibility into post-click behaviour the Search Console report cannot show.
- Document the source of every number you cite internally — date, sample, surface.
What to do before changing content strategy
The most common mistake I see right now is overreacting to a single dashboard movement: a sharp week-over-week dip in impressions becomes a “we lost to AI Overviews” narrative inside three days. Two principles that reliably keep this in check:
- Build a repeatable observation window. Define a minimum cycle (often 14 days for blog content, longer for low-volume terms) and don’t act before it closes. A new AI view in Search Console is most useful as a longitudinal series, not a snapshot.
- Prioritise where business value and measurement confidence overlap. If you can’t reliably measure a query cluster, don’t make it your first rewrite target — even if the impressions look impressive.
For supporting technical surfaces, lean on what Google publishes directly. The Search Central structured data documentation remains the canonical reference for what does and doesn’t influence rich-result eligibility. Structured data does not guarantee AI citations — it only makes content easier for Google’s parsers to interpret.
If you want help running this kind of measurement-first audit on your own site, that is what my marketing audit and consulting engagement is built for — we triangulate AI visibility from Search Console, manual checks, and GA4/BigQuery, and ship an evidence-led list of next moves. For longer write-ups, Insights tracks how the AI surface keeps moving, and AI Lab collects the prompts and skills I reuse during these reviews.
Frequently Asked Questions
Is the new Search Console generative AI report available to every site?
Not yet. Google said it is rolling these reports out to a subset of websites first to test them and gather feedback before making them widely available. If you don’t see them in your property today, you are not doing anything wrong.
Does the new report mean Search Console fully separates AI Overview clicks from ordinary search clicks?
No. The announcement describes a dedicated view for visibility inside AI features. Google explicitly says the same data also remains inside the overall performance report, so the new view is an additional lens, not a clean split.
Can I use the report to prove I lost traffic to AI Overviews?
Not on its own. Impressions and pages from the new view show visibility patterns inside AI features. To attribute a change in business outcomes to AI Overviews specifically, you would need to combine it with the existing Performance report, manual SERP checks, and downstream behaviour data such as GA4 raw events in BigQuery.
What is the relationship between the new report and the AI Overview measurement studies people cite?
The academic measurement studies (for example Xu et al., 2026) are sample-based research on Google Search; numbers like a 13.7% activation rate apply to that sample, not to every site or every query. The new Search Console report is your site’s own first-party data from Google. The two complement each other but are not interchangeable.
Should I rewrite content based on what I see in the new report?
Only after isolating the variable. Watch at least one full weekly cycle, annotate concurrent changes, and verify your AI visibility hypothesis manually. The dashboard tells you something is moving — your evidence note tells you whether it is worth rewriting for.