Organic traffic is falling for a lot of businesses that have done nothing wrong. Their rankings hold and their content is sound, but sessions decline anyway. The cause is structural: what counts as a search result has changed. A growing share of searches now end on the results page itself, answered by an AI Overview or a generated summary, with no click to the open web.
That is what zero-click search means. The phrase gets used loosely, so it is worth being precise about. The searches still happen. Fewer of them produce a visit.
What zero-click search is
A zero-click search is one that ends without a visit to any website. The user asks, the results page answers, and the session closes there. The answer might come from an AI Overview, a featured snippet, a knowledge panel or a map pack. Sometimes the user simply refines the query. In every case, no publisher receives the click.
The distinction that matters commercially is between demand and traffic. Search demand has not declined. People are asking search engines more than ever, and asking AI assistants questions they never typed into a search box. What is shrinking is the click supply: the share of those questions that turns into a website visit. The path from demand to traffic is closing, one query category at a time.
Most of the alarm about zero-click search treats it as an audience problem. The audience is still there and still buying. The problem is distribution: the answer now reaches the buyer without passing through your website.
The numbers, and where they come from
The scale of the shift has been measured. This is a field thick with invented statistics, so the figures below come with their sources.
The broadest figure comes from clickstream analysis published by SparkToro, which found that 68% of US Google searches in the first four months of 2026 ended without a single click, up from roughly 60% across 2024. Two out of three searches now resolve inside Google’s own surface.
The role of AI answers in that shift is measured separately. Pew Research Center, analysing the real browsing behaviour of a panel of US adults, found that when Google showed an AI-generated summary, users clicked a traditional search result in 8% of visits, against 15% on results pages without one. The sources cited inside the summary received a click in just 1% of visits. The generated answer absorbs the click, and the citation links recover very little of it.
These are US figures, and clickstream panels have known limitations. No credible dataset points the other way. The mechanism behind the numbers is an answer good enough that clicking becomes optional, and it is now working through Australian results pages too.
Who loses traffic first
The decline is uneven. Which part of a site loses traffic matters more than the headline number.
Informational content loses first, because a generated answer resolves informational queries completely. Definitions, explanations, how-tos and comparisons have answers that fit inside a paragraph, and a paragraph is what an AI Overview produces. Publishers and content-led sites built on informational traffic are losing the top of their funnel. The pages losing the most are often their best ones, because a page that answered the question well is the page the machine now answers from.
Some traffic is insulated, at least for now. Transactional queries still need a website, because the answer to “buy X” is a checkout. Branded and navigational queries survive, because the user is asking for a specific destination. Anything an answer cannot replicate keeps its reason to be visited: tools, calculators, configurators, logged-in experiences, and original data the reader wants to inspect in full.
For most businesses, the insulated traffic is the traffic they already owned. The informational traffic being absorbed is the traffic that introduced new buyers, the top of the funnel where someone discovered a brand before they knew its name. That is the layer AI answers are taking, and it fed everything below it.
The traffic that is growing
The same shift is opening a smaller channel: visits referred by the AI tools themselves.
Adobe measured it first at scale, drawing on more than one trillion visits to US retail sites: traffic referred from generative AI sources grew 1,200% between mid-2024 and early 2025. The channel kept growing after that. TechCrunch, reporting subsequent Adobe Analytics data, put the growth at 393% year on year for the first quarter of 2026. As of March 2026, AI-referred visitors converted 42% better than non-AI traffic and generated 37% more revenue per visit.
The base is still small relative to search, and the percentages flatter it. What matters is the quality of the visit. A visitor who arrives from an AI assistant has already had their questions answered and their shortlist formed inside the conversation. They arrive later in the journey and closer to a decision. The trade is fewer visits that are each worth more.
The common mistake: defending traffic
Faced with declining sessions, most organisations do one of two things. They publish more content to win back the clicks, or they consider blocking AI crawlers to stop being summarised. Both defend the wrong asset.
Traffic was always a proxy. What mattered was being present at the moment a buyer asked a question in your category, and for twenty years the click and that presence were the same event, so optimising for one optimised for the other. Zero-click search separates them. The question still gets asked and the presence still gets allocated. The click no longer travels with it.
Defending the proxy fails on its own terms. Publishing more informational content wins a larger share of a shrinking click pool, and it feeds the system doing the shrinking. Blocking the crawlers removes a brand from the answers without slowing the shift to answers. The buyer still asks, and a competitor becomes the response. The thing worth defending is presence.
What presence without the click looks like
If the click is no longer the unit of visibility, presence has to be defined without it. It takes three forms.
Citation. When an AI Overview, ChatGPT, Perplexity or Copilot answers a question in your category, your pages are among the sources the answer is drawn from and linked to. Citation is the closest thing to ranking, and it is earned through a defined body of work. How that selection happens, and what it rewards, is covered in how AI Overviews choose which brands to cite and in whether the model will cite you.
Mention. Models name and describe brands even when they do not link to them. An accurate description inside an answer the buyer never clicks away from still shapes the shortlist. This is also where inaccuracy does quiet damage, because a wrong description is repeated at scale with no analytics trail.
Downstream demand. Presence in answers shows up later as branded search, direct visits and the small, high-converting AI-referral segment. The buyer read the answer, remembered the name and arrived without a trackable path. The visit still happens, but nothing records where it came from.
None of the three appears in an organic sessions report. They still affect who gets chosen.
The measurement problem this creates
This creates a reporting problem. Organic traffic as a headline KPI now measures a shrinking channel, so a business can hold or grow its presence in front of buyers while its traffic chart points down. A leadership team that treats the sessions line as the health of organic search will draw the wrong conclusion.
The fix is to read traffic alongside the presence signals: whether the engines cite and describe the brand when asked the buying questions, what is happening to branded search and direct visits, and how the AI-referred segment converts. The first of those can be checked internally. Auditing your own AI search visibility sets out the method.
Where this settles
Zero-click search is not a temporary interface fashion. It is what search looks like once the results page gives an answer, and every engine including Google is moving that way because users prefer it. The click belonged to one era’s interface. The question, and the commercial value of being its answer, will outlast it.
That changes the job. The work now is to make sure that when the engines assemble an answer in your category, your brand is retrieved, cited and described accurately inside it. That is the discipline of AI search optimisation. Businesses that start early reach those positions before the questions in their category are settled.