What Pew measured
Pew analyzed March 2025 Google searches from 900 U.S. panelists against result pages collected in April. Its 22 July report notes that summaries can change over time.
In that dataset, traditional-result clicks followed 8% of visits with an AI summary and 15% without one. A cited-summary link was the next action in 1% of visits with a summary. These are event shares for the stated panel and period, not publisher payment rates.
Source notes: Google users are less likely to click on links when an AI summary appears in the results · What Web Browsing Data Can Tell Us About Americans’ Use of AI
What a publisher could infer
The study supports a disciplined traffic question: pages that appear in results with summaries may need separate monitoring from pages that do not. A publisher can segment search landing pages by query type, position, device, country, and date, then compare qualified visits and on-site outcomes with the period before a changed result presentation. That is an investigation design, not proof of causation.
It also makes a case for tracking the work after the click. A lower referral count may not have the same commercial meaning for a newsletter signup, a paid guide, an ad-supported page, and a sponsor landing page. Revenue reporting should record the denominator, the source, and the action that actually creates value rather than treating every lost visit as equivalent.
Why the headline cannot become a forecast
This observational panel study cannot establish what every user saw or forecast an individual publisher’s results. Our reading: a traffic model should keep the observed study pattern separate from assumptions about a different audience and period.
Keep the 8% and 15% figures with their method. Repeating only the difference turns a bounded behavioral comparison into a universal claim about search.
A practical question
Which search-led pages actually fund a business outcome, and what is their baseline conversion after arrival? Build that list before interpreting an aggregate traffic chart. If a meaningful change appears, preserve search-query and landing-page evidence, note result-page conditions where observable, and compare the same outcome over a stated window. Keep annotation separate from the traffic export so a later reader can distinguish evidence from interpretation.
Sources & limits
The findings describe a tracked U.S. panel’s Google searches in March 2025. They are not a causal estimate or a forecast of any publisher’s search traffic or revenue.
- Google users are less likely to click on links when an AI summary appears in the results
Pew Research Center original analysis · Source published 22 July 2025 · Primary source checked · 16 September 2026 - What Web Browsing Data Can Tell Us About Americans’ Use of AI
Pew Research Center methodology report · Source published 23 May 2025 · Primary source checked · 16 September 2026
Source claims and editorial judgments remain separate. Send a correction with the passage and supporting evidence.
