Search Console finally separates AI visibility from everything else. On June 3, 2026, Google rolled out dedicated Generative AI performance reports for Search and Discover — the feature it had been testing with a subset of site owners for months — giving website owners a way to see which pages show up inside AI Overviews and AI Mode without guessing from blended totals. It is a genuinely useful diagnostic addition. It is not, however, a new scoreboard, and treating it like one is the fastest way to draw the wrong conclusions from it.

The report gives you AI search impressions by page, country, device, and date — telling you that a link to your content appeared somewhere inside a generative response. It does not tell you whether anyone read it, whether the citation mattered, or whether the appearance did anything for the business. Here is what the data actually contains, how Google counts it, and how to use it without over-claiming what it proves.

What the new report shows — and what it leaves out

The Search report currently counts impressions from AI Overviews and AI Mode. Experimental Search Labs features are excluded, and generative visibility in Discover gets its own separate report. Google is rolling this out gradually to a subset of site owners while it gathers feedback, so not everyone has access yet.

You can segment by page, country, device, and date. That is the full list. The report does not currently include queries, clicks, click-through rate, average position, citation placement, the specific passage Google used, or any conversion data. Google says it is evaluating what else to add, but for now this is an impressions-only view.

Alongside the reporting, Google is also testing a control that lets site owners include or exclude their content from AI Overviews, AI Mode, and Discover’s generative features, separate from ordinary search results. Inclusion is the default. Before flipping that switch, understand the trade: opting out forfeits both the impressions and whatever traffic those features might be sending. Use the new report to establish a baseline first, rather than reaching for the off switch on instinct. See also: why GSC impressions are up but traffic is falling.

How Google actually counts an “AI impression”

Google defines an impression as an instance where a link to your site is shown to a user inside a generative feature. The catch is that the counting method changes depending on the level of aggregation you’re looking at.

In the property-level chart, two URLs from the same site appearing in a single AI response count as one impression. In the page-level table, each of those URLs can register its own impression. So if an AI Overview cites both a product page and a supporting research page from your domain, the chart might show one impression while the page table shows two. Adding up page-level numbers will not reliably reproduce the property-level total — that’s not a bug, it’s two different dimensions being counted differently.

The counting rules also vary by surface. In AI Overviews, a link generally has to be scrolled or expanded into view before it counts. In AI Mode, a follow-up question is treated as a new query, and links surfaced in that follow-up response can generate additional impressions on top of the original ones.

Do not merge this with your organic impression numbers

Both metrics share the word “impression,” but they describe different user experiences. A conventional search impression usually means a discrete listing was presented for consideration. In a generative response, the product is the synthesized answer itself — the supporting link might be prominent, buried among several citations, hidden until expansion, or absent entirely from the visible response.

That difference matters for reporting. Blending generative and organic impressions into one “total search visibility” number, or calculating a combined click-through rate across both, produces a figure that technically exists but doesn’t mean what it appears to mean. If you’re already rethinking how you frame AI visibility to stakeholders, the discipline of keeping metric types separate is central to why this new Search Console data can become a trap for marketers who reach for a single headline number too quickly.

Patterns worth digging into

The real value of the report isn’t the raw impression count — it’s comparing generative visibility against how the same pages perform in conventional search. A few patterns stand out.

Strong organic visibility, weak AI visibility. A page can rank well and pull solid organic impressions while barely showing up in the generative report. That’s not automatically a problem — not every query triggers an AI response, and not every page type is a good synthesis candidate. But it’s worth checking whether the page answers questions directly, whether headings are descriptive, whether passages stand alone without needing surrounding context, and whether key information is buried in tabs, images, or JavaScript that a crawler struggles to extract cleanly.

Modest organic visibility, strong AI visibility. This is often the more interesting pattern. A page with unremarkable conventional traffic can pull a disproportionate share of generative impressions because it contains a clean definition, statistic, or comparison that’s easy to lift. These pages are worth studying first — not because one URL reveals a secret ranking factor, but because they show what Google finds easiest to reuse: direct answers near the top of a section, strong heading structure, original data, and language that mirrors how people actually phrase questions.

Visibility concentrated in a handful of pages. If most of your generative impressions trace back to a small set of URLs, group them by template, topic, author, and content type. Often one template exposes content cleanly in HTML while another buries it behind scripts or tabs.

Movement after content revisions. The report can help you track whether substantive edits correspond to changes in generative visibility over time — but resist declaring victory after a single heading change and a bump three days later. Demand shifts, competing sources, and Google’s own systems can all move in the same window. A sustained increase after a real revision is useful evidence. It’s still a signal, not proof.

A workflow for using the data without over-reading it

You don’t need new tooling to get value out of this report — you need exports, consistent date ranges, and some restraint about what the numbers actually prove.

1. Export AI-visible URLs over a meaningful date range. A few days of rollout data won’t tell you much.
2. Pull conventional search data for the same URLs and dates — organic impressions, clicks, CTR, average position, top queries — from the standard Performance report, which still carries the query and click detail the generative report lacks.
3. Categorize the pages by type, topic, intent, template, author, and last substantial revision. A list of URLs and counts is inventory, not analysis.
4. Investigate the outliers. Look at what one overperforming template or author is doing differently in structure, answer placement, and internal linking.
5. Layer in analytics separately to identify actual AI referral traffic and conversions, without assuming impressions translate directly into sessions.
6. Track trends, not daily swings. Google flags this data as preliminary and subject to change; monthly comparisons are far more reliable than day-to-day noise.

This pairs well with running structured tests to prove AI search is actually working for your content rather than relying on impressions alone. And if opting a site out is on the table, review what opting out of Google’s AI search features actually means first — the switch exists, but so does the gap between having it and having the data to decide how to use it.

What to keep off the executive dashboard

The moment a large new number labeled “AI” appears, someone will want it enlarged and colored green at the top of a report. Resist blending it with anything it doesn’t belong next to. Avoid reporting a combined generative-plus-organic impression total, a blended CTR, self-reported “AI market share” based only on your own numbers, revenue attributed to impressions without supporting evidence, or page-level totals mislabeled as property-level exposure.

A more defensible dashboard tracks generative impressions over time, the number of pages receiving them, how AI-visible pages compare to organically visible ones, and any AI referral traffic identified separately through analytics.

Frequently asked questions

What does the new Search Console generative AI report actually measure?

It measures AI search impressions — how often a link to your site appeared inside AI Overviews or AI Mode responses — segmented by page, country, device, and date. It does not include queries, clicks, CTR, or citation placement.

Why doesn’t page-level data add up to the property-level total?

They use different aggregation rules. The property-level chart counts multiple URLs from the same site cited in one response as a single impression, while the page table can count an impression for each URL individually.

Should I combine AI impressions with organic impressions in reporting?

No. They represent different user experiences and are counted under different rules. Blending them into one “total visibility” figure or a combined CTR produces a number that looks meaningful but isn’t reliably comparable.

Should I opt my site out of AI Overviews and AI Mode?

Only after establishing a baseline with this report. Opting out removes your content from generative features entirely, along with any traffic those features send — it’s a real trade-off, not a free privacy toggle.

Search Console‘s generative report is a diagnostic lens, not a final score. Its value comes from comparison: AI-visible pages against organically visible ones, topics and templates against each other, visibility before and after real revisions, and exposure against whatever business outcomes you can actually trace. The useful question isn’t whether the number went up — it’s what the pattern tells you about how Google is using your content.