Google has repeatedly argued that AI Overviews improve search, deliver higher-quality traffic to publishers, and create more opportunities for websites to be discovered and clicked. New independent research challenges all three of these claims. The findings show that AI Overviews suppress clicks to the open web without producing more valuable visits, and that citations within the answers do little to make up the difference. 

New experimental research from Saharsh Agarwal and Ananya Sen tests these claims directly. The findings show no measurable improvement in user satisfaction or in the quality of visits to external websites. They also show that without AI Overviews, Google Search sends more traffic to external websites and, even when AI Overviews cite their sources, those links recover only a small fraction of the outbound traffic lost.   

The researchers recruited 1,065 U.S. desktop users and randomly assigned them to different versions of Google Search. Participants conducted their own searches as usual. One group saw Google normally, including an AI Overview whenever Google generated one. For another group, a browser extension removed the AIO but left traditional search results in place. The researchers then compared behavior across the two groups. 

AI Overviews reduce traffic and do not improve quality 

When researchers remove the AI Overview in search, users click through to external websites more often. Outbound organic clicks rise from 0.37 to 0.62 per search, an increase of about 68%. In other words, AIOs reduce publisher clicks by nearly 40%.  

The effect also showed up in zero-click behavior. When AI Overviews appeared, 73% of searches ended without a click to an external site, compared with 54% when the AI Overview was removed. In other words, AI Overviews weren’t simply changing which results people clicked; they were making an external click substantially less likely. 

The additional traffic generated without the AI Overview appears comparable in quality to traffic generated with AIOs. Researchers find no significant differences in time spent on the destination site, bounce rates, or how often users return to Google. 

A post-research survey of participants also tests Google’s claim that people have “higher satisfaction with their search results” when they use AI Overviews. The researchers find no measurable improvement. Satisfaction with Google Search averages 4.0 out of 5 both with and without an AIO. Ratings for information quality and ease of finding information also remain essentially unchanged. 

Informational searches are the hardest hit 

The experiment also helps identify where the traffic impact is most likely to occur. During the research period, Google generated AIOs for about 41% of all searches. But their presence varied considerably depending on the type of search. 

Informational searches face the greatest exposure. These queries seek information or answers rather than a specific website or an action, such as making a purchase. Informational queries account for 71% of searches, and Google generates an AIO for 53% of them. 

AIOs appear on 15% of transactional searches, where users intend to take an action, such as making a purchase. They appear on just 6% of navigational searches, where users are trying to reach a particular website or page.  

The location where AIOs appear on the search page also matters. More than 87% appear above all organic results, and the researchers find that the decline in outbound traffic is concentrated in those searches. When an AIO appears at the top of the page, removing it increases organic clicks by 88%. When it appears lower on the page, the researchers find no measurable effect. 

AIO citations recover little traffic for websites 

The experiment also confirms what the digital media industry has already seen in referral data: links within AI Overviews do little to offset the reduction in outbound traffic. Among outbound clicks on search pages where an AIO appears, only 7.1% come from links within the AI-generated answer. The remaining 92.9% come from traditional organic results on the same search page. 

Importantly, these percentages represent a smaller pool of traffic because AIOs reduce the number of users clicking to external websites. AIO citations therefore generate relatively little referral traffic compared with traditional organic search results. 

For publishers, the findings directly challenge Google’s justification for AI Overviews. Google says AIOs increase search satisfaction and produce higher-quality clicks. This independent experiment finds no evidence supporting either claim. Without AIOs, user traffic to external websites increases, while the quality of those visits remains comparable.