Newsroom artificial intelligence adoption is broad but uneven. In a 2025 WAN-IFRA member survey, 49% of publishers said they were just starting to use AI, while 30% were increasing use and 13% described their use as advanced. The same survey found 8% were not using AI yet. These results describe early member-survey conditions, not a single global adoption rate.
Contents
- How widely newsrooms are using AI
- Where publishers are deploying AI
- Audience-facing products and plans
- Governance, oversight and tool choices
- Value, barriers and workforce effects
- What audiences accept
How widely newsrooms are using AI
The 2025 WAN-IFRA survey shows a newsroom sector still moving through stages of experimentation. Almost half of surveyed publishers, 49%, said they were just starting to use AI. Another 30% said they were increasing their use, and 13% classified their use as advanced. The remaining reported position in this survey was 8% of publishers not using AI yet. Source: WAN-IFRA, “Survey finds gaps between expectations of publishers and current results of using AI”, 2025.
UK journalist responses provide a different view because they measure individual frequency rather than publisher maturity. Among UK journalists surveyed in 2025:
- 27% used AI for at least one journalistic task daily.
- 29% used AI for at least one journalistic task one to four times per week.
- 16% used AI for at least one journalistic task one to three times per month.
- 11% used AI for at least one journalistic task less than monthly.
- 16% had never worked with AI on a journalistic task.
Source: Reuters Institute, “AI adoption by UK journalists and their newsrooms: surveying applications, approaches, and attitudes”, 2025.
These measures should not be combined as if they describe the same population. The WAN-IFRA figures concern publishers and organisational self-assessments; the Reuters Institute figures concern surveyed UK journalists and the frequency of at least one task.
Where publishers are deploying AI
Publisher priorities have moved beyond a narrow focus on writing. In the Reuters Institute’s 2024 planning survey, 56% of surveyed news executives rated back-end automation, including transcription and copyediting, as a key AI priority. Recommender systems followed at 37%, human-supervised content creation at 28%, commercial AI uses at 27%, coding at 25%, and AI-supported newsgathering at 22%. Source: Reuters Institute, “Journalism, media, and technology trends and predictions 2024”, 2024 planning survey.
The 2025 planning survey reported higher levels of importance across several operational and editorial uses. Back-end automation was considered very important by 60% of publisher respondents. AI for personalisation and recommendation was important to 80%, while AI for content creation was important to 77%. Source: Reuters Institute, “Journalism, media, and technology trends and predictions 2025”, 2025 planning survey.
| AI area | 2024 key priority | 2025 reported importance |
|---|---|---|
| Back-end automation | 56% | 60% |
| Recommender or personalisation systems | 37% | 80% |
| Content creation | 28% for human-supervised creation | 77% |
The labels and question wording differ between the planning surveys, so the table is a compact comparison of reported survey measures rather than a calculated trend. It does not establish that every category rose by the displayed difference.
The 2026 planning survey points to continuing interest in practical infrastructure. In that survey, 97% of publisher respondents considered back-end automation important to their organisations, 82% considered AI newsgathering cases important, and 81% considered AI-assisted coding and product development important. The same survey found that 75% expected agentic tools to have a large or very large impact on the news industry in the near future. These are planning-survey responses and expectations, not completed adoption outcomes. Source: Reuters Institute, “Journalism, media, and technology trends and predictions 2026”, 2026 planning survey.
Audience-facing products and plans
Publisher plans for audience products show a preference for adapting existing journalism into additional formats and services. In the 2025 publisher survey, 75% planned to explore turning text articles into audio. AI summaries at the top of stories were under exploration by 70%, and 65% planned to explore translating news articles into different languages with AI.
More interactive uses were also present. Some 56% planned to explore AI chatbots or search interfaces for audiences, while 36% planned to experiment with turning text stories into video. Source: Reuters Institute, “Journalism, media, and technology trends and predictions 2025”, 2025 planning survey.
These percentages describe plans to explore or experiment, not the share of newsrooms that had launched each product. That distinction matters for interpreting newsroom artificial intelligence adoption statistics: a planned audio conversion, summary feature or chatbot is evidence of intended direction, not evidence of routine audience use.
Governance, oversight and tool choices
The 2025 UK journalist survey indicates that formal safeguards were present in many outlets, but not uniformly across every issue. Overall, 60% of journalists said their main outlet had established AI protocols or guidelines for at least one surveyed issue. Specific protocol coverage was reported as follows:
- 44% said their outlet had protocols for human oversight and control.
- 43% said their outlet had protocols for AI data privacy and security.
- 42% said their outlet had protocols for AI transparency.
- 27% said their outlet had protocols for AI bias and fairness.
- 32% said their main outlet provided training on AI technologies.
Source: Reuters Institute, “AI adoption by UK journalists and their newsrooms: surveying applications, approaches, and attitudes”, 2025.
The same UK survey described how outlets sourced their tools. A majority, 57%, said their outlet used only third-party AI tools. Another 34% said their outlet used a combination of third-party and in-house tools, while 9% said their outlet used only AI tools developed in-house. The figures describe reported tool arrangements among surveyed journalists; they do not measure the quality or effectiveness of those tools.
The pattern places human oversight, privacy, transparency and fairness in the same operational conversation as procurement. A newsroom can rely on external products while still needing internal rules for review, data handling and disclosure. The survey percentages show the reported presence of those arrangements, not whether every use case was covered.
Value, barriers and workforce effects
Publishers in the 2025 WAN-IFRA survey identified several obstacles to AI adoption. Lack of technical talent was an AI challenge for 60% of publishers. Editorial controls and accuracy were challenges for 56%, while ethical and legal concerns affected 50%. Integration with existing systems was identified by 48%, cost and resource constraints by 43%, and uncertainty about return on investment by 34%.
Source: WAN-IFRA, “Survey finds gaps between expectations of publishers and current results of using AI”, 2025.
The same survey separates hoped-for value from value reported as already helpful. Time savings were important to 84% of respondents, but 59% said AI was currently helpful for that outcome. Productivity increases were sought by 82%, while 47% said AI had already helped deliver efficiency. These paired measures show a gap between strategic expectation and current reported results; they do not provide a calculated return on investment.
Workforce effects were reported in the Reuters Institute’s 2026 planning survey. There, 67% said AI had not reduced jobs so far, 16% reported slight staff reductions, and 9% reported added roles or costs. The categories do not account for every possible response, and the survey wording describes reported organisational experience up to the survey period. Source: Reuters Institute, “Journalism, media, and technology trends and predictions 2026”, 2026 planning survey.
What audiences accept
Adoption inside a newsroom does not automatically translate into public comfort. In a six-country audience survey conducted in 2025, 12% were comfortable with news made entirely by AI. Comfort rose to 21% for AI-made news with a human in the loop and to 43% for a human-led process using some AI help. Entirely human-made news had the highest reported comfort level at 62%.
Source: Reuters Institute, “Generative AI and news report 2025: How people think about AI’s role in journalism and society”, 2025 six-country audience survey.
Audience comfort also varied by the task assigned to AI. In the same 2025 survey, comfort was 55% for AI editing of spelling and grammar and 53% for AI translation. It was lower for rewriting content for different audiences, at 30%, and for creating a realistic image when no photo exists, at 26%. Artificial presenters or authors had a comfort level of 19%.
The figures suggest a clear distinction in public response between assistance and authorship, but they should remain tied to the six-country survey population and its 2025 measurement period. They are not a worldwide approval rate, and they do not measure actual audience behaviour. For journalism organisations, the most relevant comparison is between the operational uses publishers are prioritising and the kinds of AI involvement audiences say they find acceptable.