Reuters: 10% Get News From AI, but Only 4% Click Through

The Reuters Institute's 2026 Digital News Report just quantified the threat publishers feared but could not measure: one in ten people worldwide now gets news from AI chatbots every week, yet only 4% of the public ever clicks through to the original reporting. ChatGPT, Gemini, and Perplexity are becoming answer engines — not gateways. The journalism business built on referral traffic just lost the link.

The numbers:

• Weekly AI chatbot news use hit 10% globally, up from 7% in 2025 — a 43% relative jump in twelve months.
• Just 1% call chatbots their main news source. For now.
• Among people under 35, usage is 16% — triple the rate of the oldest age group.
• South Korea leads at 14%. The UK trails at 4%, the lowest market surveyed.

Who's using them:

• Adoption concentrates among people already obsessed with news: 18% of "news lovers" use chatbots weekly, versus 7% of casual consumers.
• Political extremes over-index too — 16% on the far left, 15% on the far right.
• These are not disengaged users looking for headlines. They are power readers who found a faster interface.

What they're doing:

• 42% ask follow-up questions — the top use case across 45 markets.
• 35% want the latest headlines; 34% want summaries; 33% want help judging source reliability.
• The demand is interpretation, not distribution. Chatbots compete with explainers and newsletters — not homepages.

The click-through collapse:

• Just 4% of all respondents say they always or often click through from AI chatbots to original sources.
• Search engines: 19%. Social media: 17%.
• Even among chatbot users who do click, the motivation skews toward verification — checking whether the bot got it right — not hunger for more detail. The answer already satisfied them.
• South Korea is the outlier: 8% of the full population clicks through from AI, and 56% of chatbot users do — still below search's 44% user click-through rate.

The trust paradox:

• Only 20% of the public trusts news from AI chatbots — versus 37% who trust most news most of the time.
• But 44% of actual chatbot users say they trust AI-sourced news. Use creates belief.
• Dr. Amy Ross Arguedas, who authored the Reuters chapter, found that trust and usage track together more tightly for AI than for social media. People choose chatbots deliberately; they stumble into Facebook.

Why publishers should panic (quietly):

• Social media and video networks already overtook news websites as the most-used news source globally — 54% versus 51%.
• AI chatbots are the next intermediation layer, and they are growing faster than news use overall.
• Reuters Institute director Rasmus Kleis Nielsen warned the industry has "a small window of opportunity" to negotiate licensing and compensation before distribution dilemmas harden — and noted that "trying to drive click-throughs to article URLs may well not be the only thing to focus on."

What this is not:

• The data covers standalone chatbots — ChatGPT, Perplexity, Gemini — not embedded Google AI Overviews, which may already be compressing search referrals further.
• Self-reported click-through behavior may over- or understate reality. But the directional gap between AI (4%) and search (19%) is enormous.

Convina's view: Publishers are fighting the wrong war. Replicating chatbot summarization is a race no newsroom wins against platform companies running frontier models at scale. The defensible asset is what AI cannot easily fabricate: named sources, institutional accountability, reporting that forces institutions to act, and provenance readers can verify. Every organization — not just media — should read this data as an operating warning. If your customers, employees, or citizens form beliefs inside answer engines that never link back to your disclosures, policies, or primary records, you do not have a marketing problem. You have a provenance problem.