AI referral traffic is an incomplete measure of how AI affects publisher discovery. A Scrunch study published on August 13, 2026 found that only about 1.1% of publisher visits occurring after an AI chat carried an identifiable AI-provider referrer. Most subsequent visits arrived through channels analytics platforms would classify differently, particularly direct navigation.
That finding matters because publishers, advertisers, and AI tooling teams often rely on referral dashboards to assess AI's audience impact. Those dashboards record the final click, not the full sequence in which a reader consults an AI system, later searches for a publisher, or navigates directly to its site. In that model, AI can influence a visit without appearing as its measurable source.
Scrunch based its analysis on millions of news-related searches and AI conversations from February through June 2026. Its privacy-safe, opt-in panel linked AI interactions to subsequent web activity, creating a broader view of the journey than conventional last-click reporting. The company details its methodology and findings in Scrunch's analysis of AI and news discovery.
The central result is not that AI produces no publisher traffic. It is that a visible AI referrer represents a small fraction of visits that occur after an AI conversation. Roughly three-quarters of those post-chat publisher visits came through direct navigation, about 9% through traditional search, and the remainder through indirect referrals.
A reader may ask a conversational AI for information, see a publisher named in an answer, and subsequently open that publisher's site directly or search for it. The analytics system seeing the later visit generally records the final route, not the earlier AI interaction. That limitation makes referral reports useful for measuring tracked clicks, but insufficient for measuring AI-assisted discovery.
| Measure | Observed result in the Scrunch study | Attribution implication |
|---|---|---|
| AI-referrer visits after an AI chat | About 1.1% of publisher visits | Tracked AI clicks represent a limited share of subsequent activity |
| Direct visits after an AI chat | Roughly three-quarters of publisher visits | AI-influenced journeys may be classified as direct traffic |
| Traditional-search visits after an AI chat | About 9% of publisher visits | A later search can obscure an earlier AI interaction |
| News searches with AI Overviews | Publisher clicks around 20%, versus about 30% without an overlay | Raw click-rate differences require query-level context |
The distinction has governance consequences. If leaders treat referral data as a complete record, they may understate AI's role in brand exposure, audience acquisition, and the value created when publisher content is surfaced in AI answers. Conversely, a broader attribution model should not turn correlation into proof. A post-chat visit can be observed, but the study's results do not make every later direct or search visit exclusively attributable to AI.
Scrunch separates generative overlays in search results, called AI Overviews, from conversational systems such as ChatGPT. This matters because the two surfaces sit at different points in a reader's path to a publisher.
AI Overviews appeared in about one in four news searches in the study. Searches with an overview produced publisher clicks around 20% of the time, compared with roughly 30% for searches without one. Taken alone, that ten-point gap could suggest a significant reduction in referrals.
However, when Scrunch held the same query constant, the gap narrowed to roughly two percentage points. The result suggests that the broader difference is substantially connected to the context and types of searches that generate an AI Overview, rather than demonstrating a universal referral decline caused by the overlay itself. For publishers, that is an important analytical constraint: aggregate comparisons can blend the impact of an interface with the characteristics of the underlying query.
The study also found that publishers named in AI answers saw sizable, context-dependent increases in visits. Major publishers showed stronger follow-on visits after being named. This supports the idea that visibility inside an answer can create a discovery effect even where the next visit does not preserve an AI referrer.
That does not eliminate the challenge for publishers. A mention is not equivalent to a click, and the research does not establish a single universal value for being cited by an AI system. The effect depends on context, including the publisher and the user journey. Still, it reinforces why publishers need to assess both direct referral traffic and the wider pathways that follow AI exposure.
For publishers and AI platform teams, the practical priority is to avoid treating one dashboard field as the whole audience story. A stronger measurement approach can compare AI referrals with changes in direct navigation, branded search, publisher mentions, and query context while maintaining clear governance over what those signals can and cannot prove. AI answers increasingly shape the routes through which audiences encounter brands and sources. Scalevise can help teams evaluate whether their organization is appearing accurately and competitively across those experiences with an AI Visibility and GEO assessment. This gives decision-makers a clearer basis for connecting AI presence to measurement priorities, content governance, and search strategy before they rely on incomplete referral data. Start an AI Visibility scan.
What did Scrunch find about AI referrers and publisher visits?
Scrunch found that AI referrals with an identifiable AI-provider referrer accounted for about 1.1% of publisher visits that occurred after an AI chat in its February through June 2026 panel data.
Why can referral dashboards undercount AI's influence?
Referral dashboards typically record the last click. They cannot reconstruct an earlier AI interaction when a reader later reaches a publisher through direct navigation, traditional search, or another indirect route.
How did AI Overviews affect publisher clicks in the study?
AI Overviews appeared on about one in four news searches. Publisher clicks occurred around 20% of the time with an overview and about 30% without one, although the same-query comparison narrowed the gap to roughly two percentage points.
Do publisher mentions in AI answers lead to more visits?
The study found sizable but context-dependent lifts in visits after publishers were named in AI answers, with major publishers showing stronger follow-on visits.
What does the study imply for publisher measurement?
It suggests publishers should assess AI referrals alongside direct navigation, traditional search, indirect referrals, publisher mentions, and query context rather than relying only on last-click referral reporting.
Scrunch's findings show that AI's effect on news discovery cannot be reduced to visible AI referral traffic. With only about 1.1% of post-chat publisher visits carrying an AI referrer, last-click dashboards capture a narrow slice of a broader journey. For publishers, the relevant task is to measure AI exposure and downstream audience behavior with enough context to distinguish observed traffic routes from AI's potential role in discovery.