Zero-Click Search: What the Numbers Actually Say

Zero-click search is the phrase publishers use for a simple event: someone searched, got their answer on the results page, and never visited anyone. The term is older than AI answer engines, but AI Overviews turned it from an edge case on simple lookups into the default experience on a growing share of informational queries. This page does one job the coverage mostly skips: put the real numbers side by side, with their dates and methods, and reconcile why they disagree.

What is a zero-click search?

A zero-click search is a query that ends without a click on any result — the results page itself satisfied the intent. Google has served zero-click experiences for years through featured snippets, knowledge panels, and direct answers like weather and conversions. What changed with AI Overviews is scope: a generated summary can now answer multi-sentence informational questions that once required visiting a page, and it sits above every organic link. What an Overview is and when it appears is covered in AI Overviews explained.

How much traffic does zero-click search actually take?

The honest answer is a set of dated measurements, not one number. These are the closed set of figures we consider well-sourced, each with its period and method stamped:

One number you will notice is missing: a sitewide "X% of all searches are zero-click" figure. Estimates of that shape circulate widely, but none we can trace to a source on our list survives scrutiny across methodologies — so we do not repeat one. The measured findings above are narrower and more useful: they tell you what happens on queries where an AI summary appears.

Why do the zero-click numbers disagree?

Because the three sources measure three different things, and each is honest about its own frame. Pew watched people: a panel of US adults' real browsing, counting whether a visit that included an AI summary produced a click. Ahrefs watched pages: a large corpus of queries and top-ranking URLs, comparing CTR where Overviews appeared. Seer watched clients: CTR across the accounts it manages, before and after Overview presence.

A user panel answers "how do people behave"; a corpus answers "what happens to a top-ranked page"; client data answers "what did managed sites experience." An 8-percentage-point drop in user click rate (Pew), a 34.5% relative CTR decline (Ahrefs), and a 61–70% organic CTR loss (Seer) are not contradictions — they are different denominators pointed at the same event. Direction is unanimous: when a generated answer appears, clicks fall hard. Anyone quoting only the scariest of the three, undated, is marketing; the reconciliation is the finding.

Does being cited in the AI answer get your clicks back?

Some — and this is the most decision-relevant split in the data. Seer's 2025 client data found sites cited inside an AI Overview earned about 35% higher CTR than uncited sites on the same queries. Pew's panel found users clicked any cited source on only about 1% of visits with an AI summary. Both are true at once: the citation is a meaningful relative advantage on a shrunken base, not a restoration of pre-Overview volume.

That means the click penalty is not distributed evenly. It concentrates on pages that are neither the answer nor named in it — which is the practical argument for competing to be the cited source rather than mourning position three. Whether that competition repays the effort for your economics is worked through in are AI citations worth anything, and the structural work of earning selection is the subject of our generative engine optimization guide.

What should a publisher actually do about zero-click search?

Start by measuring your own exposure instead of adopting an industry average — the sitewide numbers above are about affected queries, and your query mix decides how much of your traffic is affected at all. In Search Console, segment your top queries by whether they are informational heads (Overview-prone) or branded and transactional (largely spared so far), and read the CTR trend per segment with dates attached.

Then be honest about business model, because zero-click is a revenue problem only for some models. If you monetize pageviews with display ads, the measured declines are structural and no optimization fully reverses them — the strategic options are covered against our own interest in is SEO dead. If you monetize being known, chosen, and contacted — lead generation, services, products — presence inside the answer retains value even when the click disappears, and the cited-vs-uncited CTR gap becomes the number worth competing for.

What we do on our own three builds reflects that split: we publish answer-first pages designed to be quotable, track AI referrals in dedicated tagging on our oldest build (live 2026-08-12), and treat citations as a measured output alongside clicks [our data]. Not because zero-click is good news — because the direction of the data is not waiting for anyone's preference.