Pew Finds Fewer Clicks When Google Shows AI Summaries. A Plan for News Sites

Publishers have argued for more than a year that Google's AI summaries reduce clicks to their sites. Last week, independent data backed them up. A Pew Research Center analysis published July 22 tracked the actual browsing of 900 US adults in March 2025 and found that users who saw an AI summary clicked on a traditional search result in 8% of visits, compared with 15% when no summary appeared.

Other findings are just as relevant for news publishers:

• Users clicked on a link inside the AI summary itself in only 1% of visits to pages with a summary.
• 18% of the 68,879 searches in the study produced an AI summary.
• Users were more likely to end their browsing session entirely after a search page with an AI summary, 26% of the time, than after one without, 16%.
• Longer searches and questions were far more likely to trigger a summary. 60% of searches phrased as questions did, compared with 8% of one- or two-word searches.
• News sites accounted for 5% of the sources linked in AI summaries, the same share as in standard results.

The study has limits: one month, US adults, and a panel of 900 people. But it measures behavior rather than opinions, and its direction matches what many newsrooms report in their analytics. It is time to plan around it.

What the numbers mean for news specifically

News is somewhat less exposed than evergreen how-to content, because a lot of news search is short and navigational: a name, an event, a team. Pew found short queries rarely trigger summaries. The exposure is highest for explanatory journalism, such as “what does the new tariff mean for car prices” or “how does the Fed decide rates.” Those question-style searches are exactly where summaries appear most, and they are often the stories newsrooms invest in most heavily.

The practical conclusion is a smaller search funnel for explainers and service journalism, and more pressure on every visit that does arrive.

Step 1: Measure search by query type

If you only watch total search referrals, you will miss the shift. Use Search Console to group queries into navigational (brand, names, events), short topical, and question-style. Track clicks and click-through rate for each group over time. You are looking for whether question-style CTR falls while navigational holds, which is the pattern Pew's data would predict.

Step 2: Make each session worth more

The most direct lever on revenue when traffic shrinks is revenue per session, and the most direct lever on that is pages per session. For news sites:

• Recirculation. Put related coverage, timelines and “what happens next” links inside the story, not only at the bottom. Readers who arrive for one explainer often want the context around it.
• Live pages. Live coverage of elections, storms and sports keeps readers on a single page for long periods. Pair it with viewability-based ad refresh that only refreshes in-view slots after a minimum time in view.
• Ad experience. Fewer, better-placed units with reserved space usually beat dense layouts. Readers who have a bad experience do not come back, and returning readers are now even more valuable.

Step 3: Grow the audiences you own

Pew's finding that people often end their session after an AI summary is a reminder that search is increasingly a destination, not a doorway. The audiences that are yours are:

• Newsletters, especially topic-specific ones that match your strongest coverage areas.
• Registered users, which also give advertisers authenticated audiences with consent.
• App and push notification users, who come directly to you for breaking news.
• Homepage habit, which is built by consistently being the best source on topics your readers care about.

Step 4: Sell what the summary cannot

An AI summary can restate facts. It cannot deliver the context advertisers pay premiums for: a trusted brand, an engaged reader, and content they know is professionally produced. News publishers have long struggled with keyword blocklists that keep brands away from hard news. Now is a good time to revisit that conversation with buyers, backed by clear content categories and brand suitability signals, so the audience you keep is valued at its real worth.

Step 5: Decide your AI access policy

The click data sharpens the question of what AI systems should be allowed to do with your content. Separate search crawlers from AI training crawlers, set a policy for each, and use network-level controls where your CDN supports them. Blocking Googlebot to avoid AI Overviews would remove you from search entirely, so that is rarely the right move.

Do not overcorrect on ad load

When traffic falls, the tempting response is to add ad units to every page. It rarely works for long. More units dilute viewability, which lowers the price buyers are willing to pay for every impression, and heavier pages load more slowly, which hurts both readers and search performance. A better approach is to test: add or remove a unit on a subset of pages, measure revenue per session and return visits over several weeks, and keep only the changes that improve both. News audiences that feel respected come back, and return visits are the traffic AI summaries cannot intercept.

What to report to leadership

1. Search clicks by query type, month over month.
2. Pages per session and revenue per session, by source.
3. Share of traffic from owned channels: newsletter, direct, app.
4. Revenue per thousand sessions, not just CPM.

AI summaries will not replace the need for original reporting, and the Pew data shows news links still appear in them. But the path from search to your site is getting narrower. News organizations that focus on session value and direct relationships will be better positioned than those waiting for the old traffic to return. HBDR works on the monetization side of that plan; the audience strategy belongs to the newsroom.