Searches with an AI summary had nearly half as many clicks - what should we measure now?

What the research found: How do AI summaries affect clicks?

U.S. research organization Pew Research Center analyzed 68,879 Google searches from 900 adults in March 2025 (published July 2025). Bain, a consulting firm, released the results of its December 2024 survey of 1,117 people in February 2025.

• Searches with an AI summary: 18% of searches studied. 8% of searches containing 1 or 2 words, 53% of searches with 10 or more words, and 60% of searches starting with a question word
• Clicks on standard search results: 8% with a summary · 15% without one
• Clicks on sources in a summary: About 1%
• Users who ended their internet session after a search: 26% with a summary · 16% without one
• Bain survey: About 80% of consumers rely on zero-click results for at least 40% of their searches. Bain estimated that search-driven visits could fall by 15% to 25% and suggested measuring impressions and AI reach instead of clicks.
• Another measurement approach (Datos): U.S. desktop behavior data showed that zero-click searches stood at 22.4% in March 2026, the lowest during the study period. Datos uses a different method from Bain’s survey and excludes mobile.
• Google’s response: Google disputed the study, citing methodological flaws and saying the search-query sample did not represent actual traffic.

Our take: Do fewer visits mean fewer patients?

Pew looked only at Google searches, did not analyze health questions separately, and did not conduct an experiment to determine whether AI summaries caused the differences. Bain’s figures are based on respondents’ self-reported answers and the company’s estimates, and the calculations have not been made public. Both studies cover the U.S., not Naver. Applying their findings to hospitals is therefore our interpretation.

In South Korea, both Naver (81.6%) and ChatGPT (54.5%) are widely used (used within the past 3 months, Opensurvey). Rather than applying U.S. figures directly, it’s more useful to track how your hospital appears in Naver and AI answers.

A drop in visits alone doesn’t tell you whether patients saw your hospital’s information. It’s time to track how your hospital appears in AI answers alongside website visits.

1. Visit numbers tell only part of the story. For 26% of searches with a summary, users ended their session without visiting another site. If patients found what they needed in the summary, that may not show up in your website analytics.
2. Pay closer attention to searches phrased as questions. AI summaries appeared far more often for searches starting with a question word than for shorter searches. U.S. findings may not apply directly to South Korea, but patients asking in full sentences about symptoms or choosing a hospital could see similar results.
3. Keep your metrics simple so you can track them consistently. Adapt Bain’s suggestions of impressions and AI reach to your hospital by checking whether it appears in answers to a set of questions - and whether the information is accurate. Consistent tracking matters more than sophisticated tools.

What is your hospital tracking?

• Do you record AI answers to a set of patient questions each month, using the same method?
• Do you track website visits alongside calls, bookings, and consultation inquiries in one place?
• Do you separately count patients who mention AI answers during consultations?

It’s hard to judge a website’s role based on a single number showing fewer visits. Track the same metrics for several months in one place, and it will be easier to see what’s actually changing.

Sources

Pew Research Center, 「Google users are less likely to click on links when an AI summary appears in the results」 (2025.7.22)

Bain & Company, 「Goodbye Clicks, Hello AI: Zero-Click Search Redefines Marketing」 (2025.2.19)

Original sources: Pew · Bain · Report on Google’s response: PPC Land · Report on Datos: PPC Land

Opensurvey, “AI 검색 트렌드 리포트 2026” (published by Opensurvey): Opensurvey Blog