Clicks Halve When an AI Summary Appears
When an AI summary appears at the top of a results page, the share of users clicking a traditional result falls from 15 percent to 8 percent — behaviour nearly cut in half. The figure comes from Pew Research Center’s analysis of the real browsing records of nine hundred users: observed conduct, not self-report, which matters because people describe what they think they do while browsing logs record what they actually did. Same engine, same users, two different worlds — pages with a summary, pages without.
Most businesses read this as “our traffic will fall”. Incomplete reading. What the data actually says is that visibility has changed floors — and the new floor has its own rules, which most competitors have not yet learned to play by.
What the Number Says
BU BÖLÜMÜN ÖZETİ
- The summary substitutes for the click
- Even the summary’s own links go unclicked
- Summaries do not appear everywhere
Three figures from the study, read together, complete the picture.
The summary substitutes for the click
The drop from 15 to 8 percent means users are finding what they need inside the summary. The same study shows sessions ending entirely far more often on summary pages: question answered, screen closed. From the user’s side this is a win, and where the user wins, the business’s job is not complaint but locating the new point of contact. The competitor is no longer the nine results below; it is the answer above.
Even the summary’s own links go unclicked
The study’s harshest line: links cited inside the summary are clicked in about 1 percent of visits. Being a source, then, guarantees no visit. Citation brings presence without traffic — and the space between those two words is the business’s new working ground: mention instead of visit, recall instead of session.
Summaries do not appear everywhere
The same analysis finds summaries triggering mostly on long, question-shaped, natural-language searches, while short and sharp queries keep the classic result list. No accident: the learner writes long, the buyer writes short. The summary swallows the information question; the decision question still clicks — which splits any site’s page inventory into two families that cannot share one prescription.
What the Headline Misses
BU BÖLÜMÜN ÖZETİ
- The loss separates by query type
- Citation is the new shopfront
- Measurement habits lag the event
Headlines count the lost traffic; nobody writes up the position to be won.
The loss separates by query type
Halved clicking is an average, and beneath the average run two different fates: definition and general-information queries surrender their clicks to the summary, while sharp-intent queries hold their behaviour. A business staring at total traffic panics; one reading query by query discovers most of what it lost was traffic that never sold. Panic economises in the wrong place; separated data shows which loss is grief and which was dead weight.
Citation is the new shopfront
A mention in the summary, even clickless, makes the brand part of the answer — a familiarity that travels to the moment of purchase. Our own panels show the local version: AI-sourced impressions climbing while clicks lag, the visibility-selection scissor now opening on a second floor.
Measurement habits lag the event
Most reports still count only clicks, while the equation has grown to three lines: did the query trigger a summary, were we cited in it, were we chosen after it. A report that cannot separate the three explains falling clicks with the wrong diagnosis — usually “rankings dropped” — and budgets the wrong medicine. A ranking that holds while clicks fall is now ordinary; the alarm belongs to decision queries only.
What a Business Should Do
BU BÖLÜMÜN ÖZETİ
- Sort traffic by type before judging it
- Sharpen the shopfront where clicks survive
- Play to be the answer’s source
Three work items follow from this data.
Sort traffic by type before judging it
Information, decision and brand queries go into separate baskets before any loss analysis. Watching only the total repeats, on the summary floor, the same single-metric trap that ranking built on the classic floor.
Sharpen the shopfront where clicks survive
Decision queries are where clicking lives on, and there the title and description are the auction floor for what remains. Rewriting those two lines is the fastest-returning job in this set for exactly that reason.
Play to be the answer’s source
On information queries the goal has moved: citation, not click. The page must answer the question in its first screen, concretely and quotably. This will not bring the traffic back; it keeps the brand inside the answer — and a brand remembered inside the answer collects its click later, at the decision.
The study itself is at Pew Research Center’s publication.
The summary halved the click but could not halve the intent; the learner stays on screen, the decider still knocks on the door.