A zero-click search is a search that produces no visit: the user found what they needed on the results page itself and left without clicking. The phenomenon predates AI — answer boxes, weather, opening hours, currency conversion and the local pack already produced this behaviour. Generative answers simply accelerated it a notch.
The scale is measured. According to Similarweb, the share of Google searches ending without any click rose from 56% in May 2024 to 69% in May 2025, a thirteen-point jump that tracks the AI Overviews rollout timeline. The Pew Research Center study, run across 68,879 searches, separately measured a click-through rate to a classic result of 8% when an AI summary is present, against 15% when it is not.
The headline figure deserves suspicion, though, because it blends situations that have nothing in common. Zero-click does not land evenly. It hits informational queries hard — how-to, definitions, tutorials — which a summary replaces outright. It barely touches transactional queries: nobody books a table, orders or makes an appointment inside a synthesis paragraph. And it does not touch navigational queries at all, when someone types your brand name.
For a local business that distinction changes everything: if your traffic comes from action-intent searches, your real exposure is low. The risk has moved elsewhere — no longer the loss of clicks, but absence from the shortlist. When an answer names only two or three businesses, the stake becomes the citation itself. That is the whole point of AI search optimization.
Common mistakes
- Taking the headline figure as your own risk: 69% zero-click does not mean you will lose 69% of traffic. It depends entirely on the intent behind the queries that bring you visitors.
- Concluding SEO is over: generative answers draw first on pages that already rank well. Search optimization remains the entry ticket — it simply no longer suffices.
- Ignoring visibility without clicks: being cited without being clicked still builds brand awareness, and those mentions count for the models.
- Failing to record a baseline: with no before-and-after numbers, separating a zero-click effect from ordinary seasonality becomes impossible.