Since August 31, 2026, a new menu has shown up in your Search Console, and it now applies to everyone: the Search Console generative AI report is live on every property on the planet. Good news, you can finally see how often your pages surface in AI Overviews and AI Mode. Bad news, Google hands you a single metric, impressions. No clicks, no CTR, no queries, no API access. The report answers “am I showing up?” and leaves untouched the only question your client cares about: “is it bringing me traffic?”.
This guide sorts out what the report gives you, what it hides, and above all how to still produce a usable indicator this week. If you are after the full AI Overviews measurement framework, it is laid out in my guide on measuring the impact of AI Overviews on your organic traffic. Here, we focus on this new report and the gap it leaves.
What changed on August 31, 2026
The generative AI performance report was announced on June 3, 2026 on the Search Central Blog, then rolled out gradually. On August 31, Google wrapped the global rollout: every property has access, no exception and no waitlist. You will find it in Search Console, in the Performance section, under a dedicated tab for AI-generated search results, next to the classic Performance report.
Concretely, this is the first time Google isolates how your pages perform on its generative surfaces. Before, those impressions already existed but stayed buried in the global Performance report, impossible to tell apart from classic results. The Search Console generative AI report finally separates them. That is real progress, as long as you know exactly what it contains.
What the Search Console generative AI report actually contains
The report aggregates three surfaces under one label: AI Overviews (the AI answer block at the top of the results), AI Mode (the conversational search mode) and the AI features in Discover. Immediate warning: it blends them. You cannot isolate AI Overviews from AI Mode. If an article promises to split the three apart, it is lying to you.
For each page, you get the number of impressions across those surfaces, broken down by URL, country and device, with granularity from hourly to monthly. That is useful: you see which pages Google deems worthy of citing in its AI answers, and how that exposure moves over time.
Two scope limits to keep in mind. First, the data only starts on May 18, 2026, with no backfill: no year-over-year comparison will be possible before May 2027. Second, Google flagged in late August that it had fixed missing data in this report, so do not be surprised if your curves shift retroactively. And this report only measures Google. It says nothing about your presence in ChatGPT, Perplexity or Claude, which is a completely different kind of measurement covered in my guide on measuring your brand visibility in AI answers.
The 4 blind spots of the report
Here is the honest state of play, as I write this in September 2026. Check this table when you read it: this is the information most likely to go stale if Google finally opens the gates.
| Data point | Available in the generative AI report? |
|---|---|
| AI impressions (AI Overviews, AI Mode, Discover) | Yes, by URL, country and device |
| Time granularity | Yes, hourly to monthly |
| Clicks from AI surfaces | No, never isolated |
| CTR specific to AI surfaces | No |
| Queries that triggered an AI answer | No |
| API access (Search Analytics) | No, the type field has no AI value |
| Dedicated column in the BigQuery export | No |
The four blind spots that matter: no clicks, no CTR, no queries, no automation. The Search Analytics API and the BigQuery bulk export still do not expose the AI search type. The type field accepts web, image, video, news, discover and googleNews, and that is all. You cannot pull these AI impressions through an API or wire them into a pipeline. That gap is exactly what makes automation impossible today.
One mistake shows up everywhere, correct it with your clients right away: clicks from AI features are not lost. They are counted, but inside the global Performance report, mixed in with everything else. AI traffic is not invisible, it is just not isolated. Believing that this traffic vanishes is the single most common misreading of the topic.
The workaround: estimating a usable AI CTR
Since Google gives neither clicks nor CTR on AI surfaces, we are going to build one ourselves. Let me be blunt before we start: this is not a real AI CTR, it is a directional proxy. It tells you whether the gap is widening or narrowing, not a figure to carve in stone. That honesty is precisely what sets you apart from the content selling a made-up “AI CTR”.
The logic comes down to three steps.
1. Freeze the AI impressions. The report has no API, so you capture by hand. A weekly snapshot is enough: export AI impressions by URL every Monday, or wire the Looker Studio Search Console connector onto the AI results dimension to log the history automatically. The point is to freeze the data before it moves, since Google keeps retouching it.
2. Cross with organic sessions on the same scope. Take the same set of URLs and pull, on the GA4 side, the associated Organic Search sessions, or go through the Search Console BigQuery export for classic organic clicks. The exact mechanics of this bridge between GA4 and Search Console are described in my guide on SEO analyses to cross in GA4 and Search Console. If you would rather do it all in SQL, the template queries are in my BigQuery cookbook for GA4.
3. Read the gap. On the same set of pages, compare how AI impressions and organic sessions each evolve. AI exposure climbing while organic stalls or drops is the signal of capture by AI answers. You are not measuring a click-accurate CTR, you are measuring a leak trend, and that is already far more than the raw report gives you.
The minimal table to build in a sheet or a dashboard, one row per URL:
| Column | Source |
|---|---|
| URL | Search Console generative AI report |
| AI impressions (week) | Generative AI report, weekly snapshot |
| Organic sessions (week) | GA4 or GSC BigQuery export |
| Sessions / AI impressions ratio | Calculated, tracked over time |
This ratio is not an official CTR, but its week-over-week evolution is the best directional indicator available today. For the chatbot traffic side, which is a distinct leak never to be confused with amputated organic, the capture method is in my guide on tracking ChatGPT, Gemini and Claude traffic in GA4.
The AI features opt-out control
The same August rollout introduced a control that lets you remove your site from Google’s AI features. The temptation is real: if AI Overviews are stealing my clicks, why not opt out. It is almost always a bad idea, and here is why.
What the opt-out does: it removes you from Google’s AI surfaces. What it does not do: it does not give you the clicks back. You do not suddenly recover the traffic captured by AI Overviews, you simply lose an extra bit of exposure, including the kind that sometimes drives real visits. You stay in the classic blue links, but in a SERP where the AI block now dominates the top of the page. Opting out means giving up one storefront without getting the previous one back.
Do not confuse this with blocking AI crawlers, which is a different lever on a different plane. Opting out of AI features in Google Search has nothing to do with stopping AI bots from training their models on your content. These two decisions are steered separately, and I cover the second lever in my guide on which AI crawlers to block, with logs and BigQuery. Confusing the two risks cutting yourself off from one channel while thinking you are protecting another.
What to put in a client report this week
No need to wait for Google to open up clicks to ship something solid. Three metrics are enough to frame the conversation.
First, total AI impressions and their trend, pulled straight from the report: that is your presence indicator on generative surfaces. Second, the top URLs most exposed in AI, to show which pages Google cites the most. Third, the organic sessions to AI impressions ratio described above, presented honestly as a capture proxy, not a CTR. Add a line of context reminding that AI clicks are still counted in global organic, and you shut down the “my traffic is gone” panic on the spot.
FAQ
Why does the generative AI report not show clicks? Because Google does not isolate them. Clicks from AI Overviews and AI Mode do exist, but they are counted in the global Performance report with no AI label. The dedicated report only surfaces impressions.
Can you get this data through an API or in BigQuery? No. Neither the Search Analytics API nor the BigQuery bulk export exposes the AI search type. The type field stays limited to web, image, video, news, discover and googleNews. Any automation therefore goes through a manual capture or an interface connector.
Why is my history so short? The data only goes back to May 18, 2026, with no backfill. You will see no history before that date, and any year-over-year comparison will be impossible before May 2027.
Can you block your site from AI features without losing your rankings? The AI features opt-out removes you from generative surfaces without giving back the lost clicks, and it is not the same as blocking training crawlers. In almost every case, turning it on costs more than it returns: you lose one storefront without gaining another.
In short
The Search Console generative AI report is a genuine step forward for answering “am I showing up in Google’s AI answers?”. It does not yet answer “is it bringing me traffic?”, and it will not until Google opens up clicks, the API or a BigQuery column. In the meantime, freeze the AI impressions, cross them with your organic sessions, track the ratio over time, and present it for what it is, an honest directional proxy. Right now that is the only serious way to turn an impressions curve into a decision.