The Pew Data on Zero-Click Search: What Happens When an AI Summary Appears

What Pew measured, and how it measured it

Most numbers in this field come from search vendors reporting on their own tools or their own clients, which makes independence and method hard to verify. The Pew Research Center study is different in kind. It is a nonpartisan, non-industry-funded read on actual behavior, and it publishes its method openly.

Pew recruited a panel of 900 US adults who consented to have their real web browsing recorded, then analyzed 68,879 unique Google searches those panelists performed during March 2025. Of that set, 12,593 searches (about 18 percent) returned a Google AI Overview at the top of the page. Because the data is observed browsing rather than a survey of what people say they do, it sidesteps the recall bias that weakens self-reported click studies.

The design also lets Pew do the one comparison that matters here: hold the searcher constant and look at what changes when an AI summary is present versus absent. That within-panel contrast, not any single headline percentage, is what makes the finding worth reading closely.

AI Overviews and clicks: the 8 percent versus 15 percent

The headline result is a paired comparison. In searches where an AI summary appeared, users clicked a traditional organic search result in about 8 percent of cases. In searches with no AI summary, that figure was about 15 percent. Present the two numbers side by side and the effect is plain: the presence of an AI summary is associated with roughly half the click-through to the open web.

It is worth being precise about what that does and does not claim. It is a measured association across a large, representative sample, not a controlled experiment that randomly assigned summaries, so it demonstrates a strong and consistent pattern rather than isolating AI Overviews as the sole cause. But the direction is unambiguous and the sample is substantial, which is exactly why this study, and not a vendor estimate, is the anchor statistic for the zero-click conversation in 2026.

The one percent almost nobody notices

The second Pew number is the one that quietly dismantles a comforting assumption. A common hope among businesses is that even if the AI summary keeps the user on Google, being named inside that summary will still earn the click. The data says otherwise.

Across the searches that showed an AI Overview, users clicked a link cited inside the summary itself only about 1 percent of the time. Being the cited source, in this sample, almost never converted into a visit. That gap between being mentioned and being clicked is the practical reason a citation cannot be treated as if it were a visit. It is a visibility outcome, and it has to be measured as one, not counted as traffic that never arrives.

The session-ending effect

A third finding rounds out the picture. Pew recorded how often a browsing session ended after a given page. After users visited a page that had shown an AI summary, they ended their browsing session about 26 percent of the time, compared with about 16 percent after pages without one.

Read plainly, the AI summary more often satisfied the need on the spot. The user got what they came for and stopped, rather than continuing to a business website. This is the mechanism behind zero-click search stated in behavioral terms: the answer, not the list of links, increasingly ends the search. For any business whose model assumes the buyer will keep browsing until they land on a site, that assumption now carries measured risk.

Why organic traffic is the wrong yardstick

Put the three numbers together and a single conclusion follows. When roughly half the clicks leave the results page, when being cited inside a summary sends a click about one percent of the time, and when a quarter of sessions end at the summary, then a decline in organic sessions is no longer a clean signal of whether a business is winning or losing in search.

The classic scoreboard, sessions and organic click-through, was built for a world where the search engine returned a list and the click was the whole transaction. In an answer-first surface, a business can be more visible than ever inside AI answers while its organic traffic falls, or can hold its traffic while quietly vanishing from the answers its buyers now read. Traffic and answer-visibility have come apart. Continuing to judge AI-search success by traffic alone measures the wrong thing precisely.

The replacement is a visibility measure: how often, across the questions real buyers ask, a business is named and cited in the answer. That is a different yardstick from a rank and a different yardstick from a session. It has to be sampled directly, because no engine publishes it.

What the Pew data does not say

Rigor cuts both ways, and the study is easy to over-read. Four limits matter before anyone builds a strategy on these figures.

• It is Google only. Pew measured Google AI Overviews, not ChatGPT, Perplexity, Gemini, or Copilot. The behavior on standalone answer engines may differ and is not established by this study.
• It is a March 2025 snapshot. AI Overview coverage and behavior are moving quickly; a single month captures a moment, not a trend line, and the figures should be treated as a well-measured point rather than a permanent constant.
• It is a population average, not your business. An 8 percent click-through across a national panel says nothing specific about any one company, vertical, or set of buyer questions. It is the reason to measure your own surface, not a substitute for doing so.
• It is association, not a controlled cause. Pew observed behavior in the wild rather than randomly assigning summaries, so it shows a strong, consistent pattern; it does not isolate the AI Overview as the single cause of every lost click.

Reading it against the wider evidence

The Pew finding does not sit alone. It aligns with two peer-reviewed results that describe the surface the click is disappearing into.

First, the object of optimization has changed from a rank to a citation. The founding "Generative Engine Optimization" study built a controlled benchmark of about 10,000 queries and found that content-level moves such as adding credible citations, direct quotations, and specific statistics produced a 30 to 40 percent relative lift on its answer-visibility metric, with citing authoritative sources the single strongest lever. Whatever the click does, the thing now being competed for inside the answer is a citation.

Second, a 2025 large-scale follow-up found AI answer engines are systematically biased toward earned, third-party media over brand-owned pages, a sharper tilt than classic Google shows. That is a single emerging study rather than a settled fact, but it points the same way: winning the answer depends on signals a traffic report was never designed to capture. Taken together with Pew, the evidence says the same thing from three directions. The click is not the game anymore; the citation is, and it must be measured on its own terms.

Is search dying? A calmer reading of the AI search statistics

None of this supports the louder claim that search is dying. Pew found that roughly 82 percent of the searches in its sample did not show an AI Overview at all, and even where summaries appeared, users still clicked through in a meaningful share of cases. Classic search, the local map pack, and reputation still carry enormous weight in how buyers find and choose.

The accurate summary is narrower and more useful than either denial or doom. Attention is being reallocated across more surfaces, and on the surface where AI summaries appear, the click to the open web is measurably rarer. That is enough to retire organic traffic as the sole measure of success. It is not enough to declare the end of search. The disciplined response is to change the scoreboard, not to panic about the game.