Google's AI answers halve the clicks websites get

By Claire Bennett · 2026-08-04

Half as likely to click

The number that should be on every marketing team's whiteboard right now is 8. That is the percentage of visits in which a user clicked a traditional search result when Google served an AI-generated summary, according to a Pew Research Center browsing study published in July 2025. On pages without a summary, the click-through rate was 15%, nearly double.^[1]

The gap is not a rounding error. Google absorbs the query, returns an answer, and sends the user back to whatever they were doing before, often without a single outbound tap.

Sessions that just stop

Pew's study tracked what users did after landing on a search page with an AI summary, and the findings are blunt: sessions ended entirely 26% of the time when an AI summary was present, compared with 16% on traditional results pages.^[1] One in four visits simply concluded at the search page itself.

Clicks on links embedded inside the summary were even rarer. Users clicked a link within the AI summary in just 1% of visits, meaning a citation inside the summary is not a reliable substitute for appearing in the traditional results beneath it.^[1] The summary is a destination, not a gateway.

About 18% of Google searches in March 2025 returned an AI-generated summary, Pew found.^[1] Each percentage point of expansion compounds the traffic effect across the billions of searches Google processes daily.

The zero-click baseline was already grim

SparkToro's 2024 study of Google search behaviour found that for every 1,000 searches in the United States, only 360 clicks went to the open web, with zero-click searches accounting for 58.5% of all queries.^[3] That figure was collected before AI summaries became a routine feature of results pages. The Pew data suggests the floor has since dropped further.

Publishers, retailers and brands that built their businesses on search-driven acquisition are facing a structurally different environment, not a temporary fluctuation in an algorithm. Google has been retaining users on its own properties for years; AI summaries are accelerating that process.

Fewer visitors, but different ones

Adobe Analytics tracked traffic to United States retail websites referred from generative AI sources across the 2024 holiday season and found a 1,300% surge compared with the same period a year earlier.^[2] The absolute volume of AI-referred visits, though still small relative to search, is growing at a pace that makes it impossible to treat as a footnote.

What those visitors do when they arrive is the more interesting part. Generative AI-referred visitors browsed 12% more pages per visit and had a 23% lower bounce rate compared with visitors arriving through other channels, Adobe Analytics found.^[2] They arrived with more specific intent, spent longer looking around, and left less quickly.

Overall engagement for AI-referred visitors was 8% higher than for non-AI-referred visitors on retail sites.^[2] The pattern fits a user who has already done substantial research inside a chat interface before navigating to a specific website to check details or complete a purchase they have, in effect, already decided to make.

The conversion gap closes fast

Adobe's conversion data shows how quickly the picture is shifting. In July 2024, conversion rates for AI-referred visitors to retail sites were 43% lower than for visitors arriving through other channels. By early 2025, that gap had collapsed to 9%.^[2]

Adobe also tracked the longer arc. AI traffic to United States retail sites surged 1,324% between October 2024 and May 2026, according to separate Adobe data.^[4] Anil Chakravarthy, President of Adobe's Customer Experience Orchestration Business, said that in a world where customers often interact with an AI tool before ever reaching a business's website, visibility is everything now.^[4]

Visibility used to mean a high position in search results. It now means being the brand an AI tool mentions, recommends or summarises when a user asks a question that could end in a purchase, and the two things require meaningfully different strategies.

What the funnel looks like now

The traditional marketing funnel assumed discovery happened at the top, on a search results page, with users clicking through to websites for the research and consideration stages before converting. The Pew and Adobe data, read together, suggest that model is inverting. Discovery and consideration increasingly happen inside AI tools, and the website visit occurs, if at all, only at the bottom: confirmation and conversion.

A visitor who has already decided what they want, compared options inside a chat interface, and arrived at a product page to complete a transaction is more valuable per session than a top-of-funnel visitor who needs persuading from scratch. The Adobe conversion trend supports that reading.

Raw traffic volume becomes a poorer signal when fewer, better-qualified visitors replace many undecided ones. Session depth and conversion rate become more meaningful. The Pew study is clean evidence that AI summaries are already reshaping behaviour at scale, in real browsing sessions. The Adobe data adds the commercial dimension: the visitors who do arrive are worth more individually, and they are arriving in much larger numbers than a year ago.

Claire Bennett

Claire Bennett writes about work and workplace culture. She is interested in the gap between how organisations describe themselves and what it feels like to work inside them.