Blog

AI Overviews Cut Clicks by 40% — Now It's Proven, Not Just Suspected

By Juyel Rana · November 6, 2026 · Updated November 6, 2026

For two years, every SEO watching a traffic graph slide has had the same suspicion: the answer box at the top of Google is eating the clicks. The problem was proving it. Traffic fell, AI Overviews spread, and the two lines moved together — but two lines moving together isn't proof, and Google kept insisting total clicks were fine. Now someone ran the experiment properly, and the suspicion has a number attached to it: 39.8%.

How the experiment actually worked

Two researchers, Saharsh Agarwal of the Indian School of Business and Ananya Sen of Carnegie Mellon, recruited 1,065 US desktop Chrome users who installed a browser extension for two weeks in January and February 2026. The study was preregistered (AEA RCT Registry AEARCTR-0017393) and approved by the Indian School of Business IRB, and the paper — "The Impact of Google AI Overviews on Publisher Traffic and User Experience" — is a preliminary draft dated June 17, 2026.

Here's the part that matters. Users were randomly split in two. One group saw Google as usual. The other had AI Overviews removed in real time, with the remaining results shifting up to fill the space. Same people, same searches, same Google — the only difference was whether the Overview was there. That is what turns this from another chart into an actual cause. Earlier work could only show that clicks fell while Overviews spread. This one isolated the box itself.

The numbers, without the spin

On searches where an Overview appeared, showing it cut outbound organic clicks by 39.8%. Flip it around: take the box away and clicks rose from 0.37 to 0.62 per search. Zero-click searches went up 34.5% in relative terms, from about 54% to about 72%.

Spread across all searches — Overviews appeared on roughly 41% of queries in the sample — that works out to about 18.5% fewer outbound clicks overall. The loss was concentrated at the top of the page. Positions one, two, and three gained the most when the Overview was hidden, which is exactly what you'd expect when the thing above them answers the question first.

And here's the part that should end the "but users prefer it" argument. Nothing else moved. Clicks on sponsored links didn't change. People didn't search less often. Among the users who still clicked through, time on site, bounce rates, and returns to the results page were all the same. And in the closing survey, there was no difference in satisfaction, perceived quality, or ease of finding information. The Overview didn't make search measurably better for the people using it. It just kept them on Google.

A second study, from the University of Pennsylvania and Northeastern, ran a similar test with 1,100 volunteers over seven days in a draft dated August 2026, and found the same direction. Treat it as supporting evidence, not a second headline number — a mid-study change to Google's Overview code broke its hiding logic, so only 51.1% of Overviews in that group were actually hidden.

The other studies were already pointing here

The experiment doesn't stand alone. It just gives a cause to a pattern everyone else had already measured. Ahrefs, looking at 300,000 keywords in December 2025, found the top-ranking page's click-through rate was 58% lower when an Overview appeared. Pew, tracking 68,879 real searches in March 2025, found users clicked a traditional result on 8% of visits with an AI summary, against 15% without — and only about 1% clicked a source link inside the summary itself.

Meanwhile the box keeps spreading. Semrush tracked a fixed panel of 400 US desktop keywords from September 2025 to August 2026 and watched Overview coverage climb from 45.75% to 88%, and 97.08% when weighted by search volume. SE Ranking, in October 2026, found Overviews on 82.5% of 1,000 US brand keywords. These were observations. The experiment is what makes them a cause.

What Google says

Google's head of Search, Liz Reid, said on August 6, 2025 that total organic click volume was "relatively stable year-over-year" and dismissed third-party reports as flawed. Google has published no numbers behind that claim.

Note what the claim doesn't say. "Total volume is stable" says nothing about who gets the clicks. A stable total is perfectly consistent with publishers losing clicks while Google keeps more of the session for itself — which is exactly what this experiment measured. Until Google publishes its own figures, the only independent causal number in public is 39.8%.

So what do you measure now?

If an Overview can take two-fifths of the clicks on the queries it appears for, a ranking report on its own stops telling you what you need to know. Position one with an Overview above it is a different asset from position one without one. The metric that matters now is how often your pages show up inside the answer — citations, share of answers, whatever you call it — alongside the clicks that still come through.

That also changes what "being visible" means. In the citation data behind this discussion, 62.2% of cited brand pages did not rank in the organic top ten for the same search, and 61% of citations came from sources a brand manages or can publish on. Ranking and being cited are two different games now, and a page can win one while losing the other.

A few things that help citation odds, based on what's been observed rather than promised: fresh dates on pages that deserve them, an opening that answers the question directly before the detail starts, and tables where a comparison or a set of figures is the point. None of that guarantees a citation. But a page written as a clear, quotable answer is the kind of page an Overview can lift from, and a rambling page that buries its point is not.

The caveats, because they matter

This is a working paper, not peer-reviewed yet. The sample was US desktop Chrome users who agreed to install a tracking extension, so mobile behaviour may differ. And 39.8% applies to queries where an Overview triggers — not to a site's total search traffic, which depends on how often Overviews appear for that site's queries in the first place.

None of that weakens the core finding. It just sets its edges. Within those edges, the answer to the question SEOs have been asking for two years is now measured: the box takes the clicks, users gain nothing measurable in return, and the clicks don't come back as ad clicks either.

Keep reading on heyjuyel.com

Sources

Want help getting cited?

If your rankings look fine but the clicks aren't coming, that's the gap I work on. Get a free SEO audit and I'll look at where your pages show up — in the results and in the AI answers.