How to Estimate AI Mode Clicks When Google Search Console Won't Tell You

Google finally gave us dedicated AI reporting in Search Console, then left out the one column everyone wanted.

The Search Generative AI performance reports launched on June 3, 2026 (UK first, then worldwide as of August 31, 2026). They show how often your pages appear in AI Overviews, AI Mode, and generative AI features in Discover. What they don’t show: clicks, CTR, or the queries that triggered those appearances.

So you can now prove you’re visible in AI Mode. You still can’t prove it’s worth anything.

This post covers what the report actually gives you, and four estimation methods that get you a defensible click number anyway, including the exact workflow I’d run this week, in order, with the failure modes of each method stated plainly.

What the generative AI report shows (and the gap)

Before estimating anything, know precisely what you have to work with.

Dimension	Available in gen-AI report?	Notes
Impressions	Yes	The only metric provided
Clicks	No	Bundled into the overall performance report, unlabeled
CTR	No	Can’t compute without clicks
Queries	No	The biggest gap for content planning
Page	Yes	Which URLs appeared in AI answers
Country	Yes	Geographic split of AI visibility
Device	Yes (Search only)	Not available for Discover
Date	Yes	Hourly to monthly granularity
Surface	Yes	AI Overviews, AI Mode, Discover gen-AI, reported as a group of AI features

One detail that trips people up: these impressions are not new data. Google confirmed AI impressions were already counted inside your overall performance report. The new report just breaks them out. If you add gen-AI impressions to “regular” impressions, you’ll double-count. I walked through the report itself in more detail in our GSC AI performance report guide.

The click data exists. Google has it, and clicks from AI features flow into your overall performance report totals. It’s just not labeled. That’s the crack every estimation method below works through.

Method 1: Subtract your way to an AI click ceiling

What it does: Uses the overall report vs. the search-type filters to bound how many of your clicks could be coming from AI surfaces.

Because AI impressions and clicks are folded into the overall totals, you can approach the number by subtraction:

1. Open the overall performance report for a date range (last 28 days).
2. Note total clicks and impressions.
3. Open the generative AI report for the same range. Note AI impressions.
4. Compute your AI impression share: AI impressions divided by total impressions.
5. Apply a discounted CTR assumption to that share. AI surfaces click through far worse than blue links. Seer Interactive’s study of 25.1 million impressions found organic CTR on queries with AI Overviews averaged 0.61%, versus 1.76% without (Search Engine Land’s coverage).

So a workable estimate is:

Estimated AI clicks ≈ AI impressions × 0.5–0.8%
(use the low end if you're rarely cited, the high end if you're a named citation)

Worked example from this site: aiseoshift.com’s last-90-day export shows roughly 9,700 total impressions. If the gen-AI report attributed, say, 2,000 of those to AI features, the estimate would be 2,000 × 0.61%, so about 12 clicks. A ceiling, not a promise.

Limitation: You’re borrowing an industry-average CTR, not measuring your own. Treat the output as an order of magnitude, not a KPI. It answers “is AI Mode sending me dozens of clicks or basically none?” which, honestly, is the decision-relevant question.

Method 2: GA4 landing-page cross-reference

What it does: Finds sessions that GSC’s web-search data can’t explain.

AI Mode clicks arrive as ordinary google / organic sessions in GA4; there’s no separate referrer. But you can triangulate:

1. In GA4, pull organic Google sessions by landing page for the last 28 days.
2. In GSC, pull web-search clicks by page for the same window.
3. Flag pages where GA4 organic sessions meaningfully exceed GSC web clicks (beyond the normal 10–20% discrepancy the two systems always have from timezone, sampling, and consent-mode gaps).
4. Cross-check the flagged pages against the gen-AI report’s page list. A page with high AI impressions and an unexplained session surplus is very likely earning AI Mode clicks.

We covered the GA4 setup side, channel groups, landing-page exploration, and annotation habits included, in GA4 for AI SEO tracking.

Limitation: The GA4/GSC discrepancy is noisy at low volumes. Below about 100 sessions per page per month, the noise swamps the signal. Run it on your top 20 pages only.

Method 3: Hourly impression patterns

What it does: Uses the gen-AI report’s hourly granularity, the one genuinely new capability it shipped with, to identify when AI surfaces show you, then looks for matching session bumps.

AI Mode usage skews toward research-heavy periods (working hours, weekday evenings). If your gen-AI impressions spike 2 to 4pm on weekdays:

1. Export hourly gen-AI impressions for two weeks.
2. Export hourly GA4 organic sessions for the same pages.
3. Look for correlated movement that doesn’t appear in your web-search click curve.

Correlation across enough days is evidence those AI impressions convert to visits.

Limitation: This is the most labor-intensive method and only works with steady traffic. It’s best used to confirm a suspicion from Method 2, not to generate the estimate on its own.

Method 4: Branded search lift as a proxy

What it does: Measures the second-order effect. People who saw you in an AI answer, didn’t click, and searched your brand later.

This matters because most AI exposure never produces a click at all; the value shows up as brand recall. That’s the core argument of zero-click strategy, and it’s measurable:

1. In GSC, filter queries containing your brand name. Chart weekly branded impressions for 6 months.
2. Overlay your gen-AI impression trend from the new report.
3. A sustained branded-query rise that begins after your AI visibility rises, with no campaign, PR, or social spike to explain it, is attributable lift.

Seer’s research found brands actually cited in AI Overviews earned 35% more organic clicks than brands absent from them, so visibility does route demand back to you. Some of it travels through this branded path.

Limitation: Attribution here is circumstantial by nature. Control for anything else that touched brand demand in the window (launches, newsletters, a viral post).

The workflow to run this week

Day one (30 minutes): Method 1. Get your AI impression share and the ceiling estimate. If AI impressions are under about 5% of your total, stop here. Your estimation effort is worth more elsewhere.

Day two (1 hour): Method 2 on your top 20 pages. Flag surplus pages, cross-reference against the gen-AI report’s page list.

Ongoing (10 minutes a week): Log AI impressions, estimated clicks, and branded impressions in a sheet, weekly. Method 4 only works if you have the trendline six months from now. Fold the numbers into the broader scorecard from how to measure AI visibility.

Skip Method 3 unless a specific page’s surplus demands an explanation.

What to do with the estimate

Two honest outcomes:

The number is tiny. For most sites in 2026, it will be. That’s a finding, not a failure. It tells you AI Mode is a visibility channel rather than a traffic channel, and your KPIs should shift toward impression share and citation presence rather than clicks. Set expectations with stakeholders accordingly.

The number is real. Then you know which pages earn it (Method 2 told you), and those page formats, usually direct-answer structures with data or tools, are what you replicate.

Either way you’ve replaced “Google won’t tell us” with a number you can defend, refresh monthly, and act on.