Liz Reid, head of search at Google, told Bloomberg’s Odd Lots podcast that AI previews reduce “bounce clicks” on publishers’ pages, continuing an argument she has made in public appearances since last year.
Reid appeared on the April 23 episode of Odd lots. Hosts Joe Weisenthal and Tracy Alloway asked how AI insights affect publisher traffic and ad revenue.
What Reid said
Reid described what she called “bounce clicks” as the category of clicks that AI insights reduce.
She said users who quickly click and return to search no longer need to visit the page because they get information from the preview. Those who want to read longer always click. She acknowledged that there were fewer ad clicks for some queries, but said the increase in query volume made up for that. The argument aligns with arguments Reid has made in other public appearances.
The model
Reid published a Google blog post in August stating that the volume of organic clicks from Google Search to websites was “relatively stable” year over year and that “quality clicks”, defined as visits that users do not click through quickly, had increased.
In October Interview with the Wall Street Journalshe explicitly used the phrase “bounced clicks” and said ad revenue with AI previews had been relatively stable.
Bloomberg’s appearance repeats the same basic argument made by Reid in August, describing some lost clicks as low-value visits where users would have quickly returned to search.
What Reid didn’t say
In none of these three appearances did Reid provide supporting data.
His August blog post didn’t include any charts, percentages, or year-over-year comparisons. On Bloomberg, she told Weisenthal and Alloway that Google checks whether people are searching more often as one of its key signals, without providing numbers.
Weisenthal and Alloway asked questions about traffic and monetization, but the interview did not include follow-up questions asking for evidence for Reid’s explanation.
Google has not publicly shared data allowing outside observers to test this distinction.
What independent data shows
Chartbeat data published in the Reuters Institute Journalism and Technology Trends and Forecasts Report 2026 found that Google search traffic from global publishers fell by about a third. Google Discover referrals fell 21% year over year across more than 2,500 publisher websites.
Analysis of Seer Interactive found that the organic click-through rate for queries with AI insights fell from 1.76% in 2024 to 0.61% in 2025, a drop of 61%. Seer noted that these queries tend to be information searches that have historically had lower CTRs.
Pew Research Center study of 68,000 real search queries found that users clicked on results 8% of the time when AI previews appeared, compared to 15% when they did not.
Digital Content Next, a trade body whose members include The New York Times, Condé Nast and Vox, reported a median 10% year-over-year decline in Google search rankings among 19 member publishers between May and June 2025. Jason Kint, CEO of DCN said at the time that member data offered “ground truth” about what was happening to publishers’ traffic.
Why it matters
Reid’s description of “bounce clicks” answers a question raised by the data, but she answers it without data of its own. This should be kept in mind when evaluating any public complaints from a platform that monitors metrics.
A business owner cannot verify from Reid’s appearance on Bloomberg whether the AI insights only remove low-value clicks or whether they cover all query types. Independent data measures total clicks and click-through rates, not the subset of clicks that Reid describes as low value. While Google has internal data that separates the two, it hasn’t shared it in the eight months since the August blog post.
See also: The Data Behind Google’s AI Insights: What Sundar Pichai Won’t Tell You
Looking to the future
Reid said Google measures how often people return to search. This signal tracks Google retention. Publishers need a traffic metric, but Google hasn’t shared one. In the meantime, “bounced clicks” should be treated as a statement rather than an observation.
