Pew Research Center's analysis of March 2025 browsing by 900 U.S. adults found that Google searches showing an AI summary led to a click on a traditional result link in 8 percent of visits, against 15 percent when no summary appeared. The study, published July 22, 2025, is one of the few public measurements of how AI changes the path from a search to a website, which is where much of marketing's effect from AI is felt first.
Key takeaways
- Pew tracked 68,879 Google searches from 900 U.S. adults in March 2025, and 12,593 of them produced an AI summary (18 percent).
- With a summary, 8 percent of visits included a click on a regular result link. Without one, the figure was 15 percent.
- Links inside the summary itself drew clicks in 1 percent of visits.
- Visits with a summary ended the browsing session 26 percent of the time, versus 16 percent without one.
- Pew covers Google only and one month, and Google has publicly said total click volume to websites has been relatively stable.
How Pew measured it
In its short read of July 22, 2025, Pew drew data from members of its KnowledgePanel Digital who agreed to install an app that records their browsing. For March 1 to 31, 2025, that produced 68,879 unique Google searches. Because browsing records do not show what appeared on a results page, a third-party scraping service re-ran the searches between April 7 and 17, 2025 to see which ones produced an AI summary. Clicks were inferred from the timestamp and web address of each person's next visit. A link that appeared both in the summary and in the regular results was counted as a summary click.
Fifty-eight percent of the 900 participants ran at least one search in March that produced a summary, and 18 percent of all searches did. The authors point out that summaries change over time, so the results describe how pages looked at collection.
The numbers side by side
| Measure | Visit with AI summary | Visit without AI summary |
|---|---|---|
| Click on a traditional result link | 8 percent | 15 percent |
| Click on a link inside the summary | 1 percent | Not applicable |
| Browsing session ended on that page | 26 percent | 16 percent |
About two thirds of all searches ended either with the person browsing elsewhere on Google or leaving without clicking a result, whether or not a summary appeared. That baseline matters for marketers: most searches never produced a click even before summaries, and the study measures a difference at the margin of a habit that was already click-light.
Which searches got a summary
Query length and wording mattered. One- or two-word searches produced a summary 8 percent of the time, compared with 53 percent for searches of 10 or more words. Queries opening with a question word such as who, what, when, or why produced one 60 percent of the time, and full-sentence queries 36 percent. In practice, a short brand name search and a long how-to question run into very different results pages.
The summaries themselves were short, with a median of 67 words and a range from 7 to 369. Eighty-eight percent cited three or more sources. Reference, video, and discussion sites together accounted for 15 percent of sources cited in summaries and 17 percent of sources in standard results. Government sites with .gov addresses made up 6 percent of summary sources against 2 percent of standard ones, and news sites held 5 percent in both.
What the study cannot say
Several limits apply, and Pew states most of them. The sample covers one search engine, because the authors could not reliably identify AI summaries on others. It covers one month, and summaries were re-collected weeks later. Clicks are inferred, not observed directly. The study counts visits by U.S. adults who agreed to be tracked, and it does not report revenue, sales, or lead volume for any site. A drop from 15 percent to 8 percent in click rate per visit is therefore a statement about behavior on results pages, and it does not translate into a change in orders or inquiries without a site's own analytics.
Google has disputed the broader claim that AI features are cutting traffic. In an August 6, 2025 post, its head of Search, Liz Reid, wrote that total organic click volume from Google Search to websites has been relatively stable year over year and that average click quality has increased, defining quality clicks as those where people do not quickly return to the results. That is a company statement without published figures, and Pew's article contains no response from Google. The two are not directly comparable: Pew measured per-visit click rates in a panel, and Google described aggregate volume.
Reading the result for marketing questions
- Check which metric a report uses: clicks per search, total clicks, impressions, or session length.
- Separate searches that show a summary from those that do not, since the two behave differently.
- Ask whether a claim about AI search covers one engine or several, and which month.
- Treat panel-based findings and a site's own analytics as different evidence, and compare both before drawing conclusions about a particular business.
FAQ
Did Pew find that AI summaries cut website traffic?
Pew found lower click rates on regular result links when a summary appeared, 8 percent versus 15 percent per visit. It did not measure total traffic to any site or the effect on sales.
Do people click the links inside AI summaries?
Rarely in this data. Links inside the summary were clicked in 1 percent of visits to pages that had one.
Does this apply to Canada?
The panel consisted of U.S. adults, and Pew does not report a Canadian breakdown. Canadian results may differ.
Which searches are most likely to show a summary?
Longer searches and those phrased as questions. In the Pew data, 53 percent of searches with 10 or more words produced a summary, and 60 percent of questions starting with who, what, when, or why did.
Related reading
- profiling rules for AI-driven marketing in Canada and California
- evidence on AI's effects across areas of society
- what federal rules ask of high-impact AI
- two cases of AI benefit with public evidence
More in AI Basics.
Last verified: October 2026
Sources
- Chapekis A., Lieb A., Google users are less likely to click on links when an AI summary appears in the results, Pew Research Center, July 22, 2025
- Reid L., AI in Search is driving more queries and higher quality clicks, Google, August 6, 2025 (company statement)