Zero-click search
A zero-click search is one where the person gets what they came for on the results page or inside an AI answer, and never visits a website at all. It is the behaviour that makes answer engine optimisation necessary rather than optional.
What it actually means
Someone asks a question. The answer appears as a featured snippet, a knowledge panel, a voice reply, or a paragraph written by an AI Overview, and the question is settled. No site is visited. From the searcher's point of view nothing is missing. From the publisher's point of view the traffic simply did not arrive.
The behaviour is not new; snippets and knowledge panels have been absorbing simple queries for years. What has changed is the range of questions that can now be settled without a visit. Comparisons, recommendations and multi-source research questions used to require clicking. Increasingly they do not.
Why nobody owns the term
Unlike AEO or GEO, zero-click search has no clean origin. It emerged as descriptive industry language, used in several places at once as the phenomenon became measurable, and was popularised through repeated analysis of search behaviour, most prominently by Rand Fishkin and SparkToro.
We record it as foundational with no sole inventor assigned. That is not a gap in the research; it is what the record actually shows, and a term this widely used is more honestly described as common vocabulary than as anyone's coinage.
What it does to the business case
Under a click-based model, visibility and traffic move together: rank higher, get more visits, measure the difference. Zero-click breaks that link. Impressions can rise while sessions fall, and the analytics will read like a decline even as the brand is being seen more often than before.
This decoupling is the reason a separate discipline exists at all. If the answer is the destination, then appearing in the answer, correctly named and correctly described, is the outcome, and the visit is a bonus rather than the point.
How to measure what you cannot count
The honest position is that this is unsolved. There is no equivalent of rank tracking for AI answers, because answers vary by phrasing, by user, by model and by day, and none of the major systems publish citation data.
What practitioners do instead is sample: ask the models the questions their buyers ask, repeatedly, and record whether the brand appears and how it is described. It is closer to polling than to analytics, and anyone claiming precision here should be asked to show their method.