Study Finds Google AI Overviews Cut Website Clicks Without Boosting User Satisfaction

A new study reveals that Google’s AI-generated search summaries, known as AI Overviews, significantly reduce clicks to external websites—without improving how satisfied users feel with their search experience.

The research, conducted by scholars from the Indian School of Business and Carnegie Mellon University, was recently published as a working paper on SSRN. Authors Saharsh Agarwal and Ananya Sen describe it as the first randomized field experiment examining how AI Overviews influence real-world user behavior.

How the Experiment Was Conducted

To test the impact of AI Overviews, researchers developed a Chrome extension and recruited 1,065 participants in the United States. These users, sourced via Prolific, were regular desktop Chrome users with sufficient browsing history.

Participants were randomly divided into three groups:

• Control group: Experienced Google Search as usual
• Hide AIO group: AI Overviews were removed in real time
• AI Mode group: All searches were redirected to Google’s AI-powered mode

Each participant remained in the study for two weeks, between January and February 2026. The experiment was pre-registered with the AEA RCT Registry to ensure transparency and credibility. Interestingly, more than 95% of users in the “Hide AIO” group didn’t realize anything had changed.

Key Findings from the Study

The results highlight a major shift in user behavior:

• AI Overviews appeared in 42% of search queries
• Removing them increased outbound clicks from 0.38 to 0.61 per search
• Organic clicks dropped by 38% when AI Overviews were present
• Zero-click searches rose from 54% to 72%

The placement of AI Overviews also mattered. When shown at the top of search results (which happened 85% of the time), they had the strongest impact. Removing top-position summaries nearly doubled outbound clicks, while lower placements had minimal effect.

Interestingly, ad clicks and overall search activity stayed stable—suggesting users are substituting website visits with AI-generated answers rather than searching more or less.

No Improvement in User Experience

Despite the drop in website traffic, user satisfaction did not improve.

Participants rated their search experience on a 1-to-5 scale, evaluating factors like:

• Overall satisfaction
• Quality of information
• Ease of finding answers

The results showed almost no difference between users who saw AI Overviews and those who didn’t.

According to the researchers, AI Overviews may be redirecting traffic away from publishers without delivering clear benefits to users.

How AI Mode Performed

Users assigned to Google’s AI Mode showed:

• Lower outbound click rates
• Higher zero-click behavior
• Slightly lower satisfaction scores

However, these findings are considered exploratory. Some participants dropped out, uninstalled the extension, or found ways around the setup, which may have influenced results.

Why This Study Matters

Previous research on AI Overviews has largely relied on observational data. For example:

• Pew Research Center found users clicked 8% of the time with AI Overviews vs. 15% without
• Ahrefs reported a 58% drop in click-through rate for top-ranking pages when AI Overviews appeared

This new experiment stands out because it uses random assignment, allowing researchers to isolate cause and effect more reliably.

Meanwhile, Liz Reid has suggested AI Overviews reduce unnecessary “bounce clicks,” but hasn’t shared detailed data supporting improved user outcomes.

What’s Next

The paper is still a draft and has not yet undergone peer review. The authors plan to expand their analysis and release updated findings in the future.

As Google continues to expand AI-driven features in search, understanding their real impact on users and publishers will remain critical.

Final Takeaway

AI Overviews may be reshaping how people interact with search—but not necessarily improving the experience. While users are clicking less, they aren’t reporting better outcomes, raising important questions about the long-term balance between convenience and the open web.