Artificial intelligence is already part of routine journalism work, but adoption is uneven and workplace guidance remains limited. The Thomson Reuters Foundation’s Journalism in the AI Era survey, published in 2025, found that 81.7% of surveyed journalists use AI tools. Separate 2025 research from the Reuters Institute measured AI use among UK journalists, while its 2024 public-attitudes research examined how audiences respond to AI-assisted news.
Contents
- How widely journalists use AI
- The journalistic tasks using AI most often
- UK newsroom frequency and task patterns
- Policies, employer attitudes, and training
- How journalists say AI is changing work
- Concerns about journalism and public trust
How widely journalists use AI
The broadest adoption figure in the available research comes from the Thomson Reuters Foundation’s Journalism in the AI Era survey (2025): 81.7% of surveyed journalists said they use AI tools in their work. Frequency is substantial among users. In the same survey, 49.4% said they use AI daily and 30.6% said they use it weekly.
That leaves 18.3% who said they do not use AI. The split is useful because it shows that overall adoption is not the same as universal adoption. A large majority has some level of exposure, but nearly one in five respondents remained outside AI use when the survey was conducted.
The Reuters Institute’s 2025 study, AI adoption by UK journalists and their newsrooms, provides a narrower UK measure focused on journalistic tasks. It found that 56% of UK journalists use AI professionally at least once a week. Within that group, 27% use AI for at least one journalistic task daily and 29% use it one to four times a week.
The two surveys measure different populations and use different definitions, so their percentages should not be treated as a single trend line. The Thomson Reuters Foundation figure covers surveyed journalists’ use of AI tools generally. The Reuters Institute figure covers professional AI use among UK journalists for journalistic tasks.
The journalistic tasks using AI most often
The Thomson Reuters Foundation’s 2025 survey shows that the most common reported use was improving content, such as editing, at 55.3%. Translation followed at 51.8%, while 48.8% used AI for research and inspiration. Data analysis was reported by 42.9%, and fact-checking by 37.6%.
These results place AI most strongly in supporting and preparatory work. Editing, translation, research, analysis, and checking are all activities that can sit alongside a journalist’s reporting process. The figures do not establish how much of any task was completed by AI, nor do they establish the quality or accuracy of the resulting work.
| AI use reported by surveyed journalists | Share | Source and period |
|---|---|---|
| Improving content, such as editing | 55.3% | Thomson Reuters Foundation, 2025 |
| Translation | 51.8% | Thomson Reuters Foundation, 2025 |
| Research and inspiration | 48.8% | Thomson Reuters Foundation, 2025 |
| Data analysis | 42.9% | Thomson Reuters Foundation, 2025 |
| Fact-checking | 37.6% | Thomson Reuters Foundation, 2025 |
The UK results from the Reuters Institute’s 2025 research are more granular and use a monthly threshold for individual tasks. Transcription or captioning was the most common listed task, with 49% using AI for it at least monthly. Translation was next at 33%, followed by grammar checking or copy-editing at 30%.
Story research reached 22% at least monthly. Idea generation or brainstorming and generating parts of text articles, such as headlines, each reached 16%. Fact-checking, verification, or source assessment reached 12%, while generating first drafts of text articles reached 10%.
The lower figures for generative writing tasks are notable within this UK dataset. Only 4% said they use AI daily for generating parts of text articles such as headlines, and 4% said they use it daily for story research. The survey also recorded 6% using AI for still image generation at least monthly and 2% using it for video generation at least monthly.
UK newsroom frequency and task patterns
The Reuters Institute’s UK data shows several levels of professional use rather than a simple user/non-user divide. Alongside the 27% who use AI daily for at least one journalistic task and the 29% who use it one to four times a week, 16% use it one to three times a month. Another 11% use AI less than once a month, and 16% have never worked with AI on a journalistic task.
| Frequency of AI use for at least one UK journalistic task | Share | Source and period |
|---|---|---|
| Daily | 27% | Reuters Institute, 2025 |
| One to four times a week | 29% | Reuters Institute, 2025 |
| One to three times a month | 16% | Reuters Institute, 2025 |
| Less than once a month | 11% | Reuters Institute, 2025 |
| Never worked with AI on a journalistic task | 16% | Reuters Institute, 2025 |
The task data suggests that routine assistance is more established than fully generative production. Transcription, translation, and copy-editing each had higher monthly use than first-draft generation, image generation, or video generation. At the same time, the figures do not show whether a newsroom permits publication of AI-generated material without human editing or verification.
Policies, employer attitudes, and training
Formal policy appears much less widespread than individual use in the Thomson Reuters Foundation’s 2025 survey. Only 13% of respondents said their workplace has an official AI policy, while 79.1% said there is no clear company AI policy. The contrast indicates that many journalists were using AI in an environment where organizational expectations were not clearly defined.
Employer attitudes were mixed. The largest reported category was neutrality: 47.6% said employers are neutral about AI integration. A further 24.5% said employers are cautiously supportive, and 22.6% said employers actively encourage AI use. Another 5.3% said employers have banned or opposed AI.
Training was also uneven. Among non-users, 55.3% cited lack of awareness and knowledge as a barrier. Insufficient training and support was cited by 52.6%, lack of access to AI tools by 47.4%, and concerns about accuracy by 45.0%.
Among AI users, 57.6% said they were self-taught. Taken together, these findings describe an adoption process driven substantially by individual initiative. They do not indicate that every organization lacks training, but they do show that self-directed learning was more common than a fully employer-led pathway among the surveyed users.
How journalists say AI is changing work
In the Thomson Reuters Foundation’s 2025 survey, 52.4% said AI had impacted their work over the past 18 months. The reported extent of impact varied: 12.4% said AI had affected their work to a great extent, 40.0% to a large extent, and 25.3% somewhat. Another 9.4% said it was too early to tell AI’s impact.
These categories describe respondents’ perceptions, not a measured change in output, staffing, speed, or employment. They also refer to the 18 months before the survey, so the time window matters when comparing these results with later studies.
The Reuters Institute’s Changing Newsrooms research (2023) captured expectations from newsroom leaders rather than journalists’ reported use. In that study, 74% of Reuters Institute newsroom leaders said generative AI will help increase productivity. A smaller 21% said generative AI will fundamentally change every role in the newsroom.
The difference between these two figures is informative. A large share of leaders expected productivity gains, while a much smaller share expected every newsroom role to change fundamentally. The 2023 findings are expectations, not a forecast that has been independently confirmed by the later survey figures.
Concerns about journalism and public trust
The Thomson Reuters Foundation’s 2025 survey recorded three prominent concerns among surveyed journalists. A total of 54.3% worried about AI’s impact on creativity and original reporting. Another 51.4% worried about erosion of critical-thinking skills, and 49.0% worried about increased misinformation.
Those concerns sit alongside the practical uses reported elsewhere in the survey. AI can be used for editing, translation, research, data analysis, and fact-checking, while journalists simultaneously worry that it could weaken original reporting, critical judgment, or information quality. The percentages do not measure whether these harms occurred; they measure concern about potential impact.
Audience comfort varies with both the degree of automation and people’s awareness of AI. The Reuters Institute’s Public attitudes towards the use of AI in journalism research (2024) found that 45% of people with greater AI awareness were comfortable with news produced mostly by a human journalist with some help from AI. Among people with less AI awareness, the comparable figure was 30%.
Across the surveyed countries, 19% were comfortable with news made mostly by AI with human oversight. This is a different editorial model from mostly human-produced news with some AI assistance, so the two comfort figures should not be combined. The results indicate that audience acceptance depends on how much of the journalistic process AI is understood to perform, as well as on respondents’ level of AI awareness.
For journalism organizations, the available statistics therefore describe a field in transition: widespread tool use, concentrated in support tasks; less common use for first drafts and generated media; limited formal policy; and continuing concerns about accuracy, reporting quality, critical thinking, and misinformation. The evidence also distinguishes current reported behavior from expectations about future productivity and newsroom change.