What Pew observed after a search
Pew examined the next browsing action after Google searches made by participants in March 2025. A traditional search result was clicked after 8% of visits to results pages with an AI summary and 15% of visits to pages without one. Users ended their browsing session after 26% of pages with a summary, compared with 16% without one. These are percentages of search-page visits in each group, not percentages of all internet users.
The chart shows two different next actions on the same 0–30% scale. It does not imply that the two rows add up to 100%: people could also keep searching or navigate elsewhere. Pew separately found a click on a link cited in the AI summary after just 1% of pages that contained a summary. If a URL appeared both in a summary and a traditional result, Pew counted the click with the summary.
| Observed next action | AI summary | No AI summary |
|---|---|---|
| Clicked traditional result | 8% | 15% |
| Ended browsing session | 26% | 16% |
Scale: 0–30%
Share of Google search-page visits in each group, March 2025. The bars show two separate actions; other actions are omitted. Clicks on AI-summary citations are excluded from the traditional-result row.
Data: Pew Research Center: Google users and AI summaries, July 2025
Why the comparison needs care
AI summaries did not appear at random. Pew found them on 8% of one- or two-word searches, but 53% of searches with at least ten words. Questions beginning with words such as ‘what’ or ‘why’ produced a summary 60% of the time. Those longer, question-shaped searches may lead to different click behavior even without a summary. The 8%-versus-15% comparison therefore cannot tell us how many clicks the feature itself removed.
The study used browsing records from 900 U.S. adults who installed tracking software on selected devices. Pew analyzed 68,879 distinct Google searches and found an AI summary on 12,593 of them. It reconstructed each results page in April 2025, after the March browsing window; a search page may have changed in between. The records also miss activity in apps and on untracked devices. This is valuable observed behavior, with a defined population and time period, rather than a current worldwide click rate.
Sources: Pew: search behavior and query length · Pew: original study methodology
A practical reading for marketers
A search impression can still expose a brand when no visit follows, while a website report only records the visits that arrive. Treat impressions, clicks, on-site outcomes, and direct or branded demand as separate signals. Do not label a traffic decline ‘caused by AI Overviews’ just because it coincides with their growth.
Start with important informational queries and pages. In Search Console, compare impressions, clicks, and click-through rate over consistent periods, segmented by query and page. Annotate content changes, ranking shifts, seasonality, and search-result changes. Then inspect whether qualified visits, leads, or sales moved with the click trend. This keeps a search visibility question connected to a business outcome.
Sources: Google Search Central: Search Console performance report
What to report each month
Use a short note for each affected topic: the queries and pages included, the date range, impressions, clicks, click-through rate, average position, and the site's corresponding qualified outcomes. Record the extraction date and any known changes in query mix. A stable click count with rising impressions tells a different story from falling clicks on fewer impressions.
If the question is whether an AI feature changed clicks, a controlled study of comparable queries or another credible design is needed. For routine editorial decisions, this Pew result is a reason to watch the full path from search visibility to value, not a universal forecast that your site will lose seven percentage points of click-through rate.