Google AI Mode cuts clicks and user satisfaction: Study

Google AI Mode sharply reduced clicks to outside websites without delivering a better search experience, according to a randomized field experiment involving 1,100 Google users.

Researchers forced some participants’ Google searches into AI Mode. The share of searches that led to an external website fell by 18.8 percentage points compared with when people used Google normally.

AI Mode sent fewer users to the web. People assigned to AI Mode also conducted 0.92 fewer Google search sessions per day. The decline affected several major sources of information. The share of users who clicked news sites fell by 12.5 percentage points. The share of users who clicked Reddit fell 21.2 points, while Wikipedia fell 9.9 points.

• Researchers found a different pattern when they tried to remove Google’s AI Overviews. The no-AI treatment increased external click-through rates by an estimated 8.8 percentage points.
• That result comes with an important caveat. Google changed the HTML structure of AI Overviews during the experiment, making the researchers’ blocking mechanism less effective. The intervention successfully hid about 51% of AI Overviews. Researchers applied a statistical adjustment to estimate the effect of their actual removal.

Why we care. The findings provide experimental evidence that an AI-first Google search experience can substantially reduce traffic to publishers and other websites. For publishers that rely on search referrals to reach audiences, broader adoption of AI Mode could further reduce this traffic source.

Search satisfaction fell. The drop in traffic didn’t come with evidence that users preferred the AI-first experience. Participants assigned to AI Mode reported lower satisfaction, usefulness, sense of control, and perceived personalization than those who normally use Google. Trust in Google also declined.

• Open-ended responses were more mixed. About 34% of AI Mode users gave negative feedback, compared with 29% who gave positive feedback.
• Common complaints included less control over search, difficulty navigating directly to websites, and limited links or source diversity. Positive responses often focused on efficiency and saving time.

More users turned to Google rivals. AI Mode also increased the share of participants who used another search engine. The share using Bing, DuckDuckGo, or Yahoo increased by 11.2 percentage points compared with people who normally use Google.

• The effect was especially pronounced among heavy Google users. Participants who previously averaged more than four Google search sessions per day conducted about two fewer daily sessions when assigned to AI Mode.

About the study. Researchers recruited 1,100 U.S. Chrome users in March 2026. They observed three days of normal search behavior, then randomly assigned participants to one of three groups for seven days: normal Google; a treatment designed to hide AI Overviews and block AI Mode; or searches redirected into AI Mode. The sample skewed younger, highly educated, and politically left-leaning. The experiment also lasted only seven days.

Yes, but. Before the experiment, AI Mode accounted for just 0.6% of participants’ searches. Among those assigned to the AI Mode treatment, nearly 95% of searches went through it.

• The experiment shows what happened when AI Mode effectively became the default Google search experience for these participants. It doesn’t show what happens when searchers voluntarily choose whether to use it.

The research paper. AI in Search Reduces Publisher Referrals Without Improving User Experience: Experimental Evidence