Google introduced dedicated Search Console reports for generative AI visibility on 3 June 2026. The reports separate appearances in features such as AI Overviews, AI Mode and generative AI experiences in Discover from the broader performance view.
This is a useful measurement improvement because it lets site owners identify which pages are being surfaced and where that visibility occurs. It does not turn every AI impression into a visit, lead or ranking signal, so the report should be interpreted alongside normal search and business data.
What the new reports include
Google's launch announcement lists the following dimensions:
| Dimension | What it helps answer |
|---|---|
| Impressions | How often site URLs appeared in generative AI features. |
| Pages | Which URLs received that visibility. |
| Countries | Where the visibility occurred. |
| Devices | Which device types saw the site in Search AI features. |
| Dates | How visibility changed hourly, daily, weekly or monthly. |
Google launched separate reports for Search and Discover and said the same data remains included in the overall performance report. The initial rollout covered a subset of websites while Google gathered feedback.
What the report cannot prove on its own
An impression shows that a URL appeared. It does not show that the mention was persuasive, that the user clicked, that the visit converted or that the appearance caused a later branded search.
The launch announcement also focused on visibility dimensions rather than a complete attribution model. That makes the report valuable for discovery and trend analysis, but insufficient for judging commercial return by itself.
How to establish a useful baseline
Start with a fixed comparison period and record the Search Console timezone. Export the available rows before changing content so later analysis uses the same source and date range.
Then group pages by role rather than reviewing one combined total:
- Service and product pages that can generate enquiries.
- Blog posts that explain or interpret a topic.
- Reference resources and glossary pages.
- Branded pages such as the homepage, about pages and case studies.
For each group, compare AI-feature impressions with ordinary Search clicks, conversions and assisted journeys. A page gaining AI visibility while losing clicks may be answering the query inside the result, attracting a different audience or competing with changing demand. The report alone does not identify which explanation is correct.
A weekly review workflow
Use the dedicated report to identify material changes, then investigate them in a consistent order:
- Confirm that the date range and filters match the earlier export.
- Find the pages and countries responsible for the change.
- Check whether ordinary Search impressions and clicks moved at the same time.
- Review indexing, canonicals and page availability for technical explanations.
- Compare enquiries or other qualified outcomes before changing strategy.
This sequence keeps an AI visibility change from being mistaken for a sitewide SEO change. Our Google Search Console guide covers the wider reporting context, while AI Overviews guidance explains the feature itself.
What to improve when visibility is weak
There is no special markup that guarantees inclusion in a generative AI feature. The controllable work remains familiar: publish accurate and useful pages, make important content available in rendered HTML, use crawlable links, show clear authorship and sources, and keep technical indexing signals consistent.
If a page earns impressions but no business value, inspect whether it answers the intended question, links naturally to the next useful step and supports its claims. Do not add repetitive “AI SEO” text merely to chase a report.
The bottom line
The dedicated reports make generative AI visibility easier to observe. They are strongest as a diagnostic layer: they show where exposure occurred and help teams decide which pages deserve deeper review. Commercial conclusions still require Search clicks, analytics, conversions and customer evidence.
