Google AI Mode sends fewer clicks to external websites and makes users less satisfied with Google Search. That is the conclusion of a randomized field experiment published in August 2026 by researchers from the University of Pennsylvania and Northeastern University, directly contradicting repeated public statements from Google’s own leadership.
What did the study actually measure?
The paper, titled AI in Search Reduces Publisher Referrals Without Improving User Experience: Experimental Evidence, enrolled 1,444 participants, of whom 1,100 users’ searches were tracked across a seven-day experiment. Participants were randomly assigned to one of three search conditions, and 956 completed a post-experiment survey.
The three conditions were:
- Forced Regular Search: AI Overviews (the AI-generated summary boxes shown above traditional results) were hidden, and AI Mode searches were redirected to standard Google Search.
- Natural Google Behavior: Google worked as it normally would, including AI Overviews when Google chose to display them.
- Forced Google AI Mode: All searches were redirected into AI Mode.
Participants installed a browser extension (a small add-on in their web browser) that recorded search behavior, collected browsing history, and took snapshots of results pages. This let researchers see exactly which queries triggered AI Overviews, whether users clicked through to external websites, and whether they switched to alternative search engines.
How did AI Mode affect clicks to websites?
Users assigned to Google AI Mode were substantially less likely to click through to external websites compared to users on regular Search. The pattern held in both directions: forcing AI Mode reduced outbound clicks, and removing AI Overviews increased them, with no detectable drop in how users perceived the overall search experience.
That last point is significant. According to the paper, stripping out AI features did not make users rate their Search experience any worse. The clicks went up; the satisfaction stayed the same.
Did AI Mode improve or worsen the user experience?
It worsened it, according to this study. Users assigned to AI Mode reported lower satisfaction, reduced trust in Search, and felt results were less personally relevant. They also conducted fewer search sessions per day, the opposite of what researchers had predicted going into the experiment.
One notable data point: a significant share of users forced into Google AI Mode started searching on competitor engines, including Bing, DuckDuckGo, and Yahoo, and reported a higher intention to switch to Bing permanently. The paper states that “assignment to AI Mode Search significantly increased the fraction of users searching on a competitor engine.”
For news-related searches specifically, the study did not find a clear difference in user perception between AI Mode and standard Search. The negative effects on satisfaction and trust were concentrated in other query types.
Which users were most affected?
The effects were broadly consistent across user groups, with one exception: heavy Google Search users. People who previously used Google the most showed a sharper drop in search sessions and a more pronounced negative reaction to the AI-first experience. The researchers found that “AI Mode reduced search activity more strongly for heavy users of Search than it did for lighter users.”
What Google has claimed
Google CEO Sundar Pichai and Head of Search Liz Reid have repeatedly stated that users love AI Mode, engage with it strongly, and that it sends more traffic to the open web than ever. The August 2026 paper directly contradicts both of those claims: clicks to external sites fell, and user satisfaction fell with them.
The study also noted a technical complication worth flagging. During the experiment, 94.7% of searches in the AI Mode group were successfully routed to AI Mode. In the no-AI condition, 90% of AI Overviews were hidden on day one, but the success rate fell to 0% as the experiment progressed, because Google changed the HTML structure of AI Overviews mid-study, breaking the extension’s detection logic. Overall, 51.1% of AI Overviews were successfully hidden across the experiment’s duration.
What this means for AI-search visibility
The plain-language version: when a search engine answers users inside the results page, fewer people click through to any website, including yours. This study gives that concern a specific shape backed by a controlled experiment with real users, not just analytics estimates.
The more striking finding for brands is the behavior shift. Users pushed into Google AI Mode did not just click less, they searched less and increasingly moved to Bing, DuckDuckGo, and Yahoo. That signals a fragmentation of search behavior that goes beyond traffic loss on Google alone. Brands optimizing purely for Google visibility may now need to account for a meaningful segment of users drifting to engines where traditional click-based dynamics still apply more directly.
There is also a gap this study makes concrete: the difference between being cited as a source in an AI answer and actually receiving a visit. A brand can appear frequently in AI-generated answers (high share of model, meaning how often AI mentions you relative to competitors) while its referral traffic from those mentions approaches zero. The study does not address this distinction directly, but the click data reinforces it. Being present in an AI answer is not equivalent to being visited.
The finding that removing AI Overviews increased clicks without hurting user experience is worth treating carefully. It suggests the negative effects documented here are not inherent to AI in search broadly, but specific to how Google has designed its current AI Mode experience. Whether Google adjusts that design in response to independent academic evidence, or continues to assert that users are delighted, is the variable most worth tracking over the coming months.
Quick checks before your next content or SEO audit
- Pull your Google Search Console click-through rate data segmented by query type, then compare trends since AI Overviews became widespread on your core queries.
- Check whether your brand appears as a named source in AI Mode answers for your main topics, and whether those answers include a visible link to your site.
- Review your analytics for traffic share from Bing, DuckDuckGo, and Yahoo over the last six months to spot whether alternative engine traffic is growing.
- Identify your highest-impression, lowest-click queries and assess whether an AI Overview is appearing for those terms.
- For news or time-sensitive content, monitor whether your pieces are cited in AI answers, since the study found AI Mode had less impact on news-related user perceptions.
The research was produced by the University of Pennsylvania and Northeastern University and published in August 2026 as a draft paper. Its methodology, a randomized field experiment with real users, places it on stronger empirical ground than analytics-based studies or surveys, even accounting for the technical limitations the researchers themselves disclosed.
Sources:
Search Engine Journal, Research Shows Google AI Mode Sends Less Clicks and Is A Poor User Experience, August 2026.
AI in Search Reduces Publisher Referrals Without Improving User Experience: Experimental Evidence, University of Pennsylvania and Northeastern University, published August 2026.
This article was drafted with AI assistance from the sources cited above and checked against them before publishing. The "What this means for AI-search visibility" section is Hingewise's editorial analysis, built on the frameworks and positions we use in our GEO work. How we work: Editorial policy.