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.
Why last-click attribution misses AI's role
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.
