Search Console’s Generative AI Report: What It Measures, and the Data-Quality Problem It Can’t Show You
Google’s Generative AI performance report launched in Search Console on 3 June 2026 and, as of mid-August, has expanded from a small UK test group to effectively global availability – Google is now actively pointing site owners to it with an in-product prompt.
It gives you five things: impressions, pages, countries, devices and dates. It does not give you clicks, click-through rate, average position or queries.
But the more urgent story is not what the new report contains. It is what the underlying change has done to the report you already rely on. Because every follow-up question inside AI Mode counts as a brand-new query, your standard Performance report query table has started filling with conversational fragments – rows reading “yes”, “yes go on”, “yes, pricing”. Those fragments carry impressions, positions and clicks, and they flow into every average calculated on the query dimension.
The Generative AI report cannot show you this. It has no Queries dimension at all.
Written by Konrad Szymaniak, Founder of Szymaniak Digital, an enterprise SEO and AI search consultancy. Konrad has led search programmes for organisations including Frasers Group and Sports Direct, speaks at brightonSEO and MeasureFest, and guest lectures on enterprise AI search and agent optimisation at the University of Southampton and the University of Stirling.
Last reviewed: August 2026
Search Console’s Generative AI Report: What Happened?
On 3 June 2026, Google announced dedicated Generative AI performance reports in Search Console, sitting under the familiar Performance section. For the first time, site owners could isolate how often their URLs appeared inside generative AI features rather than having that activity folded invisibly into standard web search totals.
What the report actually contains. Five data categories: impressions, pages, countries, devices and dates, with granularity from hourly to monthly. It covers generative AI features in Google Search — specifically AI Overviews and AI Mode — as well as generative AI features in Discover. Data begins on 18 May 2026, with no historical backfill.
An important nuance that is frequently misread: this is not newly collected data. AI Overviews and AI Mode activity was already flowing into the general Performance report and remains there. The new report provides a separate, isolated view of it. Your historical totals have not changed.
The rollout has moved fast, and recently. It launched to a limited group of UK site owners, which was widely reported as a response to the Competition and Markets Authority’s requirements on Google. Through June, access widened to India, the US and Switzerland. Around 11 August 2026, Search Console began showing an in-product pop-up inviting site owners to open the new report — the first time Google has actively promoted it — and by 12 August practitioners were reporting the report live across effectively every property they had access to.
If you read earlier coverage telling you to prepare while waiting for access, check Search Console again. You very probably have it now.
Google also shipped an opt-out control. Site owners can exclude their content from AI Overviews, AI Mode and Discover AI features, taking effect from 17 June 2026. Google has confirmed the toggle is not used as a ranking signal for traditional organic results. Note the distinction that trips teams up constantly: this is a different control from Google-Extended, which governs Gemini model training and has no effect on AI Mode whatsoever.
And then the side effect surfaced.
In early August 2026, the SEO practitioner Anastasia Kourou posted on LinkedIn about strange rows appearing in her standard Search Console query table — short conversational fragments that made no sense as searches. John Mueller of Google responded confirming that Search Console includes AI Overviews and AI Mode information in the general performance report, and pointed to the relevant documentation.
That documentation states plainly that a follow-up question inside AI Mode is treated as a new query, and that all impression, position and click data from the new response is attributed to it. So when someone asks Google a question, reads the answer, and types “yes go on”, that phrase becomes a query in your Performance report — carrying impressions, a position and potentially a click.
Konrad Szymaniak, Szymaniak Digital: “This is the part that has been almost entirely missed in the coverage. Everyone rushed to analyse the shiny new report and nobody checked what the same underlying change had quietly done to the report they’ve been making decisions from for a decade. Your Performance report is no longer a list of things people searched for. It’s a list of things people searched for, plus fragments of conversations they had with a model that happened to cite you. Those are not the same data type, and they’re currently in the same table.”
Search Console’s Generative AI Report: Why Does It Matter?
1. The counting rules are not what most teams assume.
Three mechanics change how the numbers should be read, and all three are documented rather than speculative:
- An unlinked brand mention is not an impression. If an AI response names your brand without linking to you, nothing is counted. Teams celebrating “AI visibility” based on brand mentions are measuring something Search Console does not report at all.
- Hidden links only count when revealed. In AI Overviews, a link must be scrolled or expanded into view before it registers as an impression. Links behind a collapsed element count only when the user activates it.
- Follow-ups multiply impressions. In AI Mode, each follow-up question generates a new response, and links in that response can generate additional impressions. A single research conversation can therefore produce several impressions from one user.
2. Position means something different in each feature.
AI Overviews occupy a single position, and every link inside the overview inherits it. AI Mode uses standard position rules applied element by element, so a link card, an image block and a carousel each calculate independently. The consequence is that average position across a mixed set of features is not a coherent measure of visual prominence, and should not be reported as though it were.
3. Blended metrics produce numbers that look like insight and behave like noise.
Generative impressions and ordinary organic impressions are not interchangeable units. Pouring both into one bucket and calling it total search visibility produces a chart that is easy to build and impossible to act on. The same applies to computing a blended click-through rate across both — you can generate the figure, but its existence does not make it meaningful.
4. The query-table contamination affects decisions you are already making.
This is the one with genuine operational cost. Conversational fragments in the query dimension inflate row counts, distort keyword-level averages, and pollute any automated reporting, alerting or classification built on top of that table. If you run rules that flag emerging queries, cluster queries into topics, or calculate CTR by query group, those systems are now ingesting conversation debris alongside genuine search demand.
The structural problem is that only the general Performance report can show you this, because only it has a Queries dimension. The dedicated Generative AI report — the one built specifically for AI visibility — has Pages, Countries, Dates and Devices, and no query data at all. The report designed to give you AI insight is structurally incapable of surfacing the biggest AI-related data-quality issue in your account.
5. The starting date sits inside a core update.
Data begins on 18 May 2026, which falls within the May 2026 core update window. Early trend lines therefore mix generative visibility patterns with ranking movement from the update. Any baseline set on the first few weeks of data needs that caveat attached, or you will attribute core-update effects to AI behaviour.
Konrad Szymaniak, Szymaniak Digital: “I keep being asked whether the new report is good news or bad news. It’s neither — it’s a partial instrument, and the professional response to a partial instrument is to understand its calibration before you trust its readings. Impressions without clicks tell you Google used your content. They do not tell you what that use was worth. Both facts are useful. Only one of them belongs in a revenue conversation.”
Search Console’s Generative AI Report: Who Is Affected?
- Enterprise organisations with automated reporting – most operationally exposed.
The larger the reporting stack, the more places conversational query fragments can propagate silently. Query-based classification rules, automated alerting, dashboard aggregations and anomaly detection all inherit the contamination without flagging it. Enterprises are also most likely to have multiple properties at different rollout stages, producing inconsistent data availability across a single global report.
- Publishers and media – highest stakes on the opt-out decision.
This is the one group for whom opting out is a genuine strategic question rather than a reflexive mistake. Content-licensing models, paywalled material and specific legal positions can all justify it. For most other businesses it means surrendering a growing visibility channel with no ranking benefit in return. The CMA’s publisher opt-out obligations, ordered on 3 June 2026 with a nine-month implementation window, sit alongside this control and will shape how the sector uses it.
- E-commerce and retail – biggest gap between impressions and value.
Product queries frequently trigger AI features where a link is cited without a click following. Rising generative impressions against flat revenue is the expected pattern here, not a warning sign in itself – but it becomes one if nobody has explained the mechanic to the commercial team in advance.
- SaaS and B2B technology – most affected by follow-up inflation.
Long, multi-turn research conversations are exactly how software buyers use AI Mode, which means this sector sees the highest ratio of follow-up-generated impressions to genuine discrete searches. Impression growth here should be read with particular caution.
- Healthcare and regulated sectors – citation as a compliance surface.
Knowing which specific pages Google cites in AI answers is genuinely valuable for organisations with review obligations, because it identifies which content is being used to construct advice. The Pages dimension is arguably more useful to this sector than the impression counts themselves.
Search Console’s Generative AI Report: What Should Businesses Do?
1. Check whether you have the report – most people now do.
Look under Performance in Search Console for the Generative AI section, or watch for Google’s in-product prompt. Advice written before mid-August assumed you were waiting; that assumption has expired.
2. Audit your query table for conversational fragments this week.
Export your Performance report query data, sort for very short rows and conversational openers, and quantify how much of your query dimension is now conversation debris rather than search demand. This is a data-quality diagnostic, not an AI think-piece exercise, and it takes an analyst an afternoon. Then check whether any automated rule, dashboard or alert consumes that table — because if it does, it is already producing degraded output.
3. Separate generative and organic impressions permanently in your reporting.
Two distinct channels, two measurement frameworks. Never blend impressions into a single visibility figure and never compute a blended click-through rate. Document the reasoning where stakeholders will see it, not just where analysts will.
4. Cross-reference pages, because that is where the actual insight lives.
Export generative impressions and organic clicks for the same URLs and dates, then look for divergence. Pages strong in both are your genuinely non-commodity assets. High generative impressions with low organic clicks suggests content that answers a question well enough to be cited but does not create enough curiosity to earn the visit — a conversion and framing problem. Low generative impressions with high organic clicks indicates ranking-dependent traffic that is exposed as AI features expand into that query category. A list of URLs and impression counts is inventory; the comparison is the analysis.
5. Set a baseline now, and annotate it.
There is no backfill before 18 May 2026, so the earliest benchmark you can hold is the one you capture today. Annotate it with the May 2026 core update overlap so that whoever reads the trend in six months does not misattribute it.
6. Do not opt out without a specific strategic reason.
For the large majority of brands, opting out removes a growing visibility channel and returns nothing, since Google has confirmed it carries no organic ranking penalty either way. The control matters as an acknowledgement of the tension between using publisher content and compensating publishers for it. Using it defensively, without a licensing, paywall or legal rationale, is a decision to be less visible.
7. Remember what Search Console structurally cannot see.
It does not and will not measure ChatGPT, Perplexity or other external AI platforms – that requires GA4 referral tracking and dedicated visibility monitoring. It does not count unlinked brand mentions. Bing Webmaster Tools has offered its own AI performance report globally for some time, so a complete picture requires looking beyond one console.
Search Console’s Generative AI Report: What We’re Watching Next
- Whether clicks ever arrive.
This remains the single most important unshipped feature in search measurement. Google has committed to no timeline. Until AI clicks are reported as clicks, every return-on-investment conversation about AI search visibility rests on inference rather than evidence.
- Whether the Generative AI report gains a Queries dimension.
Currently the report has none, which is why it cannot surface the conversational-fragment problem. Adding queries would resolve two issues at once — giving practitioners the prompt-level view they have asked for since launch, and making the contamination visible in the report designed to explain it.
- Whether Google separates conversational fragments from genuine queries.
Google documents the current behaviour as intended, so there is no indication a change is coming. But as AI Mode usage scales — it passed one billion monthly users at I/O 2026, with query volume reported as more than doubling each quarter — the proportion of the query table made up of conversation debris will grow. At some point that becomes a reporting problem Google has to address rather than document.
- Whether the global rollout is genuinely complete.
Reports in mid-August suggested near-universal availability, though some site owners still reported not seeing it. Treat “global” as substantially rather than uniformly complete for now, and verify per property rather than assuming.
- How the regulatory picture shapes the reporting.
The report’s origins lie in the CMA’s requirements on Google, and further obligations are dated: the UK’s publisher opt-out lands roughly nine months from its 3 June 2026 order, and under an EU order issued on 16 July 2026, Google must begin sharing search data with rivals from January 2027. More of what Search Console eventually shows you is likely to be determined by regulators than by product roadmaps.
Konrad Szymaniak, Szymaniak Digital: “My advice to clients right now: spend an afternoon on the query-table audit before you spend a week on the new report. One of those tells you something you didn’t know about data you’re already using to make decisions. The other gives you an impression count you can’t yet convert into a business case. Do the diagnostic first, then build the dashboard.”
“Every follow-up question inside AI Mode lands in your Search Console query data as a brand-new search – and Google’s dedicated AI report has no query dimension to show you it is happening. Szymaniak Digital’s AI Search Visibility Audit quantifies how much of your reporting is now conversation debris, separates generative from organic performance properly, and rebuilds the measurement framework on top.”
Search Console’s Generative AI Report: Frequently Asked Questions
Do I have access to the Generative AI report yet?
Almost certainly. It launched on 3 June 2026 to a subset of UK site owners, expanded through June to India, the US and Switzerland, and by mid-August 2026 Google was surfacing an in-product pop-up pointing site owners to it, with practitioners reporting it live across effectively all their properties. Look under Performance in Search Console for the Generative AI section. If you still cannot see it, availability appears substantially rather than uniformly complete, so check again shortly.
Why are queries like “yes” and “yes go on” appearing in my Search Console?
Because they are real. Google’s documentation confirms that a follow-up question inside AI Mode is treated as a new query, with all impression, position and click data attributed to it. When someone reads an AI answer and types “yes go on”, that becomes a row in your Performance report. John Mueller confirmed this behaviour in early August 2026 when practitioners flagged the odd rows. Nothing is broken and you are not being spammed – but those fragments do flow into any average you calculate on the query dimension, which is worth auditing.
Can I see which queries trigger my AI impressions?
No. The Generative AI report has four dimensions — Pages, Countries, Dates and Devices — plus impressions. There is no Queries dimension. The general Performance report does carry query data including AI Mode and AI Overviews activity, but without a filter isolating AI features, so the two cannot be cleanly reconciled. This remains the most requested missing capability.
Does an AI mention count as an impression?
Not on its own. Google has confirmed that an unlinked brand mention in an AI response does not count as a Search Console impression. A link must be present, and in AI Overviews it must be scrolled or expanded into view before it registers. If your AI visibility narrative is built on brand mentions, you are measuring something Search Console does not report — that requires separate citation and share-of-voice monitoring.
Should I treat AI impressions the same as organic impressions?
No, and blending them is the most common mistake being made with this data. They represent different placements, different user behaviour and structurally different click rates. Report them as separate channels, never compute a blended click-through rate across both, and treat generative impressions as a signal of what Google finds citable rather than as a traffic forecast.
Should I opt out of Google’s AI features?
For most brands, no. Opting out removes your content from AI Overviews, AI Mode and Discover AI features while returning no ranking benefit — Google has confirmed the setting is not used as a ranking signal for traditional organic results. The genuine cases are publishers with content-licensing models, paywalled material or specific legal positions. Note this is a different control from Google-Extended, which governs Gemini model training only and has no effect on AI Mode.
Why does my impression count keep rising while clicks stay flat?
That is the structurally expected outcome rather than a failure. AI features are designed to answer questions in place, so citation without a click is the normal case. Follow-up questions compound it, since each one generates a new response that can produce additional impressions from the same user in the same session. Read rising impressions as evidence Google finds your content citable, then judge commercial performance on revenue per session and priority-segment lead quality instead.
Does Search Console show my visibility in ChatGPT or Perplexity?
No, and it will not. Search Console reports on Google surfaces only. Traffic arriving from external AI platforms shows up as referral traffic in GA4, and visibility inside those platforms requires dedicated monitoring. Bing Webmaster Tools has offered its own AI performance report globally, so a genuinely complete picture spans at least three sources.
About Szymaniak Digital
Szymaniak Digital is a UK-based enterprise SEO and AI search consultancy helping senior marketing leaders stay visible as search becomes conversational, agentic and personalised.
Founded by Konrad Szymaniak, the consultancy specialises in enterprise SEO strategy, Generative Engine Optimisation (GEO) and Answer Engine Optimisation (AEO) – combining traditional organic performance with visibility across AI Mode, AI Overviews, Deep Search and the wider generative search landscape.
Konrad has delivered search programmes for organisations including Frasers Group and Sports Direct, holds an MSc from the University of Southampton, speaks regularly at brightonSEO and MeasureFest, and guest lectures on enterprise AI search and agent optimisation at the University of Southampton and the University of Stirling. He contributes to Semrush, Screaming Frog, Sitebulb, Majestic and Wordtracker.
Szymaniak Digital’s Search Discovery Monitoring, Prompt Portfolio Mapping and A.G.E.N.T. Framework give enterprise clients a measurement framework built for AI search – separating generative from organic performance, quantifying reporting contamination, and tracking citation as a first-class metric alongside ranking.
