The Oxford-based Reuters Institute for the Study of Journalism did not set out to measure trust in any news. It aimed to measure something broader: what people now believe, industry-wide, about what happens when they open a news app or home page, and whether they think that change has made the experience better or worse. The response was unusually divided, even by the standards of research on AI skepticism.
What the survey actually asked
The Generative AI and News Report 2025Conducted by Felix Simon, Rasmus Kleis Nielsen and Richard Fletcher and published on October 7, 2025, the survey conducted by YouGov between June 5 and July 15, 2025, surveyed 12,217 people across Argentina, Denmark, France, Japan, the United Kingdom, and the United States.
It’s a companion piece to the Institute’s better-known annual Digital News Report, and is designed specifically to track public perception of AI’s role in journalism, rather than news consumption habits in general.
Numbers
Fifty-one percent of respondents said they believe generative AI is “always” or “often” used by news media to create content. full details of the report. It’s a belief about the industry as a whole, not a claim that any respondent made about the particular article they read that morning, but it puts them in the position of assuming that engagement with AI is now the default, not the exception, for the majority of the public surveyed. When asked if this participation has improved their news experience, only 26% said yes, 8% “better” and 18% “a little better.” The news media, along with government and politicians, was one of only three sectors in the entire survey where more people expected AI to make things worse than expected to improve them.
In fact, the question was which “better” was being asked
The report frames it more as a convenience gap than a knowledge gap: people aren’t confused about what generative AI can do for the newsroom, faster summarization, automated aggregation, translation at scale, they’re not particularly sure it’s becoming a better experience for them as any reader. Nieman Lab’s coverage of the same report, written by co-author Felix Simonnotes that even though personal use of tools like ChatGPT has nearly doubled year-over-year among the same population, this suspicion is valid. People embrace the technology as individuals, but aren’t convinced that its adoption by news outlets does them any good, a more specific and less tractable finding than the general AI concern.
Why the news sits with the government and politicians
Company news media is worth sitting in this ranking. Two other sectors where pessimism outweighs optimism are government and politicians, institutions that already have structural reasons to distrust humans regardless of AI. News media included in this group, rather than sectors such as health care or education, where the survey found more balanced or positive expectations, suggests that skepticism is not a general technology concern. This is typical of institutions that the public suspects are no longer acting in their best interests, and AI adoption is read through a preexisting lens rather than judged on its own terms.
What the void actually represents
A newsroom may deploy generative AI for entirely defensible reasons, cost pressure, speed, coverage of stories that would otherwise go unreported, and direct access to a readership that still decides the outcome will be worse, regardless of justification. This is a more difficult problem than the disclosure label addresses, as the report’s own data suggests that the belief that AI is already spreading is not coupled with any expectation that the spread is good news.
Newsrooms measuring the extent to which these tools have been significantly adopted do not discriminate between the informed public and the uninformed. They choose how to act in front of a public that already assumes that adoption has occurred and has made up its mind about what it means.
A perception that may precede experience
Nothing in the report claims that generative AI is actually “always or often” used in the news industry, but only a small majority of the public believes so. Actual adoption rates vary widely by newsroom, market, and beat, and many reports do not include any generative AI.
This gap between perceived and actual use is his own finding: mistrust of the role of technology in an institution can outweigh the institution’s actual acceptance of it, especially if the institution in question was already a low-confidence category before the technology arrived, such as government and politicians.
Correcting the record on actual usage rates may make opinion less forward than editors would like, as the report suggests that skepticism is not just a response to what AI is doing, but who is doing it.
What does this mean for how the disclosure reads
This has direct implications for disclosure policies. A newsroom that clearly labels its AI-powered content answers the question “Did AI touch this,” suggesting that the data has readers already deciding the answer regardless of the label. What the label doesn’t answer, and what the survey actually incites is negative sentiment, is whether the reader generally trusts the agency’s judgment about when and how to use the tool. Disclosure changes what readers know. This obviously does not change the structural inaccuracy of the Reuters Institute data.
One figure in the report that makes it difficult to draw any neat conclusions is the adoption rate among the same respondents: nearly double the previous year. People aren’t rejecting generative AI in their own lives, but they’re condemning it in the news. They do both at the same time, which implies that protest is never about technology in the abstract, but only about relying on an institution’s judgment on their behalf about where and how to deploy it.
Whether that changes as the technology matures or becomes more rigid depends less on newsroom disclosure policies than on whether institutions that used AI before its introduction can regain initial trust, the survey found.
