Google’s AI Overviews Cut Clicks Without Profiting Satisfaction: Report
Randomized field test finds Google’s AI overview reduces organic clicks on external websites by 38% in the queries where it appears, while self-reported search satisfaction remains almost unchanged when the abbreviations are removed.
A working paper by researchers from the Indian School of Business and Carnegie Mellon University was posted on SSRN this month. Authors Saharsh Agarwal and Ananya Sen describe it as a randomized field experiment to test how AI Overview affects user behavior in a real browsing environment.
How Temptation Works
Agarwal and Sen created a Chrome extension that randomly assigned 1,065 US participants to one of three groups. People are employed at Prolific and use Chrome on the desktop. They also had to meet minimum browsing history limits, so the sample shows more active desktop Chrome users than all Google users.
The control group saw Google Search in general. The “Hide AIO” group had an extension and removed the real-time AI Overview. The third party is redirected to Google’s AI mode for all searches. The study ran for two weeks for each participant between January and February 2026.
Researchers pre-registered trials through the AEA RCT Registry prior to data collection. Over 95% of users in the Hide AIO group saw no changes during the study.
Researcher’s Findings
AI overview appeared in 42% of queries, and removing it increased outbound clicks from 0.38 to 0.61 per search. They reduced organic outbound clicks by 38% on activated queries, with zero-click searches increasing from 54% to 72%.
The results were strongest when the AI Overview appeared at the top of the page, which happened 85% of the time. Removing the AI Overview for the top spot almost doubled the outbound clicks, but the bottom one had no effect.
Sponsored clicks and search frequency remain stable, reflecting the shift between AI Overview and organic visits.
Finding User Experience
The endline survey used a 1 to 5 Likert scale to assess participants’ search experience. Responses from the control and Hide AIO groups were nearly identical on all measures, including satisfaction, quality of information, and ease of access to information.
The researchers wrote that AI Overviews “divert traffic away from publishers without delivering a measurable improvement in user experience.“
How does the AI mode compare
Participants targeted in the AI mode had lower bounce rates, higher zero click rates, and lower bottom line satisfaction compared to the other groups.
The authors note that these results are exploratory, as higher reductions, certain extension releases, or finding workarounds may have influenced the results.
Why This Matters
Independent estimates of AI Overview’s impact on traffic were generally consistent. Pew Research found users click 8% of the time with an AI overview, compared to 15% without. Ahrefs analyzed GSC data and reported a 58% drop in click-through rates when the AI Overview appeared.
This experiment adds a different approach by randomly assigning users to see the AI Overview or not, isolating the causal effect.
Google VP Liz Reid says the AI overview has cut down on “bounce clicks,’ but doesn’t provide data to support the user benefit side. A paper by Agarwal and Sen examined a related question in a randomized design, finding no measurable change in satisfaction or ease of finding information.
Looking Forward
The paper is a draft in SSRN and is not peer-reviewed. The authors will add additional results, and we will provide an update if the findings change.
