Platforms Surpass Press and Drain Trust Downward

Global trust in information stands at 37% worldwide in 2026. Trust in social media is at 22%. And for the first time since the Reuters Institute has measured this data, digital platforms — social networks, online video, chatbots — surpass all other information sources combined, with 54% of global audiences.

These three figures, taken together, tell something more precise than a simple “crisis of trust.” They describe an arithmetic mechanism: when audiences migrate in mass from channels of medium trust to channels of low trust, global trust in information erodes — even if each channel remains stable. It is not a question of journalistic quality. It is a question of topology.

The Essentials

  • Platforms (social networks + video + AI chatbots) now represent 54% of information sources used globally, according to the Reuters Institute Digital News Report 2026 — a first.
  • Trust in social networks reaches 22%, trust in AI chatbots 20%, compared to 37% global trust in information. The news avoidance rate reaches 42% in 2026.
  • Television has lost 13 percentage points of audience since 2020, news websites 12 points.
  • The political challenge is no longer solely regulating platforms: it is financing reliable sources capable of maintaining their audience in this new configuration.

A Structural Shift, Not a Trend

The Reuters Institute has published its annual report since 2012. It today covers 48 markets, making it the largest survey ever conducted on information behavior worldwide. The 2026 shift is not a cyclical spike: it is the culmination of a constant slope.

In 2020, television remained the primary information source in the majority of covered countries. It has lost 13 points in six years. News websites, which had benefited from a rebound during the Covid period, have ceded 12 points since. These two channels are not disappearing, but they are losing their status as the main entry point into current events.

What platforms have captured is not uniform. YouTube and TikTok are driving growth, particularly among those under 35. Facebook remains dominant in emerging markets in Africa and Southeast Asia, often as the sole point of access to information for hundreds of millions of people. WhatsApp and Telegram structure increasingly private information, circulating in closed groups beyond any indexing or verification.

Generative AI chatbots are entering the information landscape. Still marginal in absolute audience, they are used by 16% of those under 35 (across all markets) as an information source each week — 17% for the 18-24 age group specifically, according to the DNR 2026. Their trust score: 20%. The lowest of all measured channels.

The Arithmetic Trap of Trust

Global trust in information is calculated as a weighted average. If the highest-rated channels lose audience and the lowest-rated ones gain it, the average declines — even without any qualitative change in each channel.

This is exactly what the 2026 data documents. Print media and public radio maintain trust scores above 45% in most European and North American markets. But their audience is declining. Conversely, social networks at 22% and chatbots at 20% capture a growing share of flows.

The consequence is counterintuitive: improving the journalistic quality of a publication that loses 10% of its readers each year is not enough to stop the decline in global trust. What matters is the ability of reliable sources to hold their position in the ecosystem — to remain visible, accessible, preferred.

This mechanism has direct political implications. We have been debating platform regulation for ten years: content moderation, algorithms, press financing. These debates are legitimate. But they do not resolve the fundamental question: even a perfectly regulated platform, stripped of its most toxic content, remains structurally a low-trust channel. Its very nature — non-editorialized content, aggregation without hierarchy, speed without verification — caps the trust that can be placed in it.

What the Report Does Not Say, but Allows Us to Calculate

The Reuters Institute measures trust, not its behavioral effects. But other studies allow us to complete the picture.

A 2024 Stanford University study on a broad panel of American adults shows that people who inform themselves primarily via social networks significantly overestimate the prevalence of extreme opinions in the population — an effect documented under the name “false polarization.” This distorted perception of social reality has measured consequences for political participation and trust in institutions.

In 2023, 36% of global respondents (across all markets) actively declared avoiding information, compared to 24% in 2017 — and this rate reached 41% specifically in the United Kingdom. The correlation with platform dominance is strong, even if causality remains debated. What the data suggests: saturation of low-signal information produces a rejection reaction that strikes both platforms and traditional press equally.

This is where the problem becomes systemic. Distrust does not remain localized on the least reliable channels: it generalizes. The 37% global trust measured in 2026 reflects not only mistrust toward TikTok or chatbots — it also reflects partial erosion of trust in sources that deserve it.

The Actors Seeking a Response

Facing this diagnosis, several strategies are at work — with uneven results.

The BBC launched its “Verify” program in 2024, a unit dedicated to real-time verification, integrated directly into its news broadcast flows. The model is based on source transparency: each sensitive fact is accompanied by a visible indicator showing how it has been verified. The impact on trust is measurable: BBC Verify subscribers report trust in information significantly above the British national average.

In France, the CSA conducted a 12-month experiment in 2025 with four television channels, imposing comparable “editorial traceability” standards — with results still partial but encouraging. France Médias Monde, which broadcasts in 180 countries, has redirected some of its funding toward Francophone African markets, where platforms have captured more than 60% of information audience in less than five years.

At the platform level itself, YouTube extended its “Information Panels” program to 40 new markets in 2025. These boxes, which signal reference sources on sensitive topics, reduced consumption of content rated “low reliability” by 15% among exposed users, according to internal data published by Google in its 2025 transparency report. It is modest. It is real.

Perhaps the most interesting case is Denmark, regularly at the top of trust in information rankings. The country combines three elements: substantial public funding for regional print press (via a distribution subsidy system), media education integrated into the school curriculum since 2018, and transparency standards imposed on recommendation algorithms of platforms operating on its territory. Global trust in information there reaches 56% — double the worldwide average.

This model is not exportable as is. But it documents one thing: trust is built by deliberate public policies, not by the sole virtue of actors.

Generative AI as an Accelerator of the Fracture

The entry of chatbots into the information ecosystem adds a new variable. Their 20% trust score is the lowest measured — but their use is growing rapidly, particularly for informational queries: international news, health, economics.

It is not the intrinsic reliability of the models that poses the problem first. It is their relationship to the source. A news article can be verified, dated, attributed to a responsible author, publicly corrected. A chatbot response bears none of these properties. It can be accurate or approximate — but the user has no reliable signal to know which.

Several labs are working on this question. The “Calibrated AI” project jointly led by MIT Media Lab and the University of Amsterdam has been testing since early 2025 interfaces that explicitly expose the sources of generated responses and their level of verifiability. Initial results suggest that better source visibility improves users’ stated trust — and, more interestingly, their ability to detect factual errors in responses.

The question is not whether generative AI will play a role in information: it already is. The question is whether it will inherit the verification and traceability practices built by journalism — or whether it will bypass them definitively. The choice is still open, and it belongs as much to designers as to regulators. The link between AI skills and the labor market follows similar logic: technical capacities generate high returns, but they only create lasting value if they rest on a framework of trust.

The Press Will Not Win at Constant Content

There is a temptation, in debates about the press crisis, to formulate the solution as a problem of format or distribution platform. Being on TikTok. Making newsletters. Adapting “storytelling.” These strategies can slow erosion. They do not reverse it.

What the Reuters Institute data suggests is more structural: the press cannot compete with platforms on their own criteria — volume, speed, immediate emotional engagement. Its differential value is elsewhere: verification, hierarchization, editorial responsibility. These are precisely the attributes that produce trust. And these are the attributes that are expensive to maintain.

In Europe, public information financing models — audiovisual licensing fees, press subsidies, additional funding for investigative support — face constant pressure. In France, the abolition of the audiovisual license fee in 2022 was replaced by a budget allocation whose durability remains subject to annual vote. In the United Kingdom, the BBC is negotiating the renewal of its charter in an unfavorable political context. In the United States, the nonprofit model — ProPublica, The Texas Tribune, The Marshall Project — progresses but remains marginal in absolute audience.

The comparison with other sectors considered public goods is instructive. Just as American biotech depends on production chains it does not control, democracy depends on an information infrastructure that finances itself less and less. The dependency is less visible, the risk less immediate — but the logic is comparable.

What the 2026 Shift Forces Us to Decide

The 54% threshold is symbolic as much as statistical. It marks the moment when platforms cease to be a complement to the information ecosystem and become its center of gravity. The topology has changed.

In this new configuration, the central question is no longer “how do we regulate platforms?” but “how do we finance credible alternatives at scale?” These two questions do not exclude each other — but the second is more urgent and less discussed.

The experiments that work — Denmark, BBC Verify, media education in schools, algorithmic transparency — have one thing in common: they require a political decision, deliberate financing, an institution that endures over time. They do not emerge spontaneously from the market, nor from the good will of platforms.

The political question posed by the 2026 report is therefore very precise: in what information ecosystem do we want to live in ten years? And are we ready to pay for building it? The answer is not in the data. It is in the budgetary arbitrations of the coming years.

Sources

  1. Reuters Institute Digital News Report 2026 — Executive Summary
  2. Google — 2025 Transparency Report (available at transparencyreport.google.com)
  3. Reuters Institute Digital News Report 2023 — data on “news avoidance” (reutersinstitute.politics.ox.ac.uk)
  4. Stanford study on false polarization and social networks, 2024 (Stanford Internet Observatory)
  5. MIT Media Lab / University of Amsterdam — “Calibrated AI” project, 2025
  6. Reuters Institute DNR 2026 – Chapter AI Chatbots
  7. Reuters Institute DNR 2026 – Denmark Page
  8. Reuters Institute DNR 2023 – Executive Summary
  9. Reuters Institute DNR 2012 – First Edition