Somewhere between 58% and 62% of Google searches now end without a click. That figure is real and it is quoted constantly, almost always to argue something it does not support.
The useful version of this story is narrower than the headline. Zero-click does not mean people stopped visiting websites. It means the click is no longer the first thing that happens — and on a growing share of queries, being named in the answer is.
What the number actually measures
A zero-click search is one that ends on the results page. The person got what they came for from a snippet, a knowledge panel, the local pack or an AI Overview, and never went to a site. Crucially, that usually means they were satisfied, not that they gave up.
The range exists because the independent clickstream studies that produce it disagree with each other by several points. They sample different panels, in different countries, counting different things as a search. Anyone quoting this to one decimal place is quoting one study and calling it a fact.
Which clicks actually went away
Not evenly distributed, and this is the part the panic usually skips. The queries losing clicks are the ones whose answer is a fact:
- Definitional. What a term means, who someone is, when something happened. The answer fits in a box and always could have.
- Conversions and calculations. Currency, units, dates, times. Google has answered these in-page for a decade.
- Local and navigational. Opening hours, a phone number, an address. The result is the destination.
What still sends traffic is anything where a person needs to compare, evaluate, or read something longer than a paragraph. A buyer choosing between four tools opens the tools. A reader wanting the argument opens the article.
So if your traffic was built on answering factual queries, this is a real problem and it started well before AI. If it was built on being useful to someone making a decision, the change is different — and it is about what happens rather than whether.
What AI Overviews changed, and what they did not
Zero-click predates generative search by years. Featured snippets, the knowledge panel and the local pack were all doing this. AI Overviews accelerated it and changed one thing that matters more than the acceleration:
The answer now names brands. A featured snippet quoted one page. A generated answer synthesises several and mentions the products in them by name. So the question stopped being “did they click through to me” and became “was I in the sentence at all” — and those have completely different remedies.
That distinction is worth sitting with. Being named in an answer someone reads and acts on is a real outcome even when nobody clicks. Being absent from it is invisible in every analytics tool you own, because an impression you never received leaves no trace.
How to tell whether it is happening to you
One signal is available and one is not, and being clear about which is which saves a lot of wasted effort.
What you can see
Open Search Console and compare impressions against clicks over twelve months or more. Impressions flat or rising while clicks fall is the signature of answers being read without visits. It is a directional signal and a good one.
What you cannot see
You cannot isolate AI Overviews as the cause. Search Console does not break out AI Overview impressions or clicks as their own appearance type — they are folded into the ordinary web performance report before you see them. Analytics cannot separate them either; a visit that started in an AI answer arrives as plain organic Google traffic.
The response that is actually available
Three things, in order of how much they matter, and none of them is about recovering the click.
- Find out whether you are named. Not whether you rank — whether your brand appears in the answers people get for your buying questions. This is observable: run the question and read the result. Doing it once tells you little; doing it weekly tells you whether you hold a position.
- Get named on the pages the answers are built from. A generated answer reaches for comparisons, roundups and discussions far more than vendor pages. If the article an engine reads every week names four competitors and not you, being added to it beats publishing a fifth page on your own site.
- Stop optimising for the clicks that are gone. Pages built to capture definitional queries are competing for traffic that is not coming back. That effort is better spent on the questions where a person still needs to choose something.
The first one costs nothing to start: run a real buyer question through real engines and read what comes back, including who got named instead of you.
The honest summary
Zero-click is real, the figure is defensible as a range, and most of what is written about it overstates the conclusion. Clicks for factual queries have been going for years. What is new is that the answer now carries brand names — which makes presence in it a thing you can lose without ever seeing it in a dashboard.
The panic is not warranted. The measurement problem is, and it is worth taking seriously: if the outcome has moved from a click to a mention, and your entire reporting stack only counts clicks, you are measuring the wrong thing carefully.
Questions people ask
What exactly is a zero-click search?
A search that ends on the results page: the person got what they needed from a snippet, a knowledge panel, an AI Overview or the local pack, and never clicked through to a website. It does not mean they left unsatisfied — usually the opposite.
Is the 58–62% figure reliable?
It is a reasonable range from independent clickstream studies, and the studies disagree with each other by several points because they measure different samples in different ways. Treat it as 'most searches, by a clear margin' rather than as a precise number — and be suspicious of anyone quoting it to one decimal place.
Does zero-click mean SEO is dead?
No. It means one outcome of SEO — the click — is no longer the only outcome, and for a growing share of queries it is not the first one. Being the source an answer is built from still requires being indexed, crawlable and clearly written, which is the same work.
Which queries lose the most clicks?
Factual and definitional ones: conversions, opening hours, what a term means, who someone is. Those were always low-value clicks. Comparative and commercial queries still send traffic, because a buyer comparing options wants to see the options themselves.
How do I tell whether zero-click is affecting me specifically?
Compare impressions against clicks in Search Console over a long window. Impressions holding steady while clicks fall is the signature. What you cannot do is isolate AI Overviews as the cause, because Search Console does not separate them — so treat it as a direction, not an attribution.