What News Organizations Built with AI, 2022–2025
This research draws on 160 AI projects from 101 news organizations launched between late 2022 and the end of 2025, making it one of the first systematic catalogs of what newsrooms built with AI. Newsrooms have spent those years using AI primarily to do what they were already doing but faster, and this efficiency-first pattern may leave the industry unprepared for the deeper disruption already changing how audiences find news. The number of newsroom AI projects grew rapidly across all the years, at least in what newsrooms have made public. Yet they do not add up to a fundamental transformation in how journalism is produced, distributed or funded.
Insights from this research can help newsroom leaders decide where to invest limited organizational capacity and which traps to avoid. They can also help people in newsrooms identify opportunities with AI for the next few years. By learning from what other newsrooms have built and what they have overlooked, news organizations can develop more resilient and sustainable operations in these disruptive times.
Six patterns run through the findings:
First, AI is primarily being used to automate routine production work. Summarization, drafting, transcription, translation and workflow automation dominate the dataset. Newsrooms are using AI tools like news production assistants to reduce repetitive tasks and increase efficiency.
Second, some AI applications expand what journalism can cover. Public meeting monitors, document analysis, investigative research tools and podcast monitoring projects allow newsrooms to process information at scales previously impossible for reporters.
Third, a small group of AI experiments treats journalism as a utility. Some organizations are building tools that help audiences navigate civic issues, such as property taxes, evictions or restaurant inspections, and answer questions about topics like immigration. While still a minority, these experiments suggest new ways of delivering journalistic value.
Fourth, public-facing AI tools pose reputational risks compared with AI in the background. Inhouse newsroom tools generally improve efficiency with limited or no public exposure. Reader-facing products such as AI summaries and news chatbots have proved far more vulnerable to errors and reputational damage.
Fifth, newsrooms are most worried about the business threat from AI and have built the least for it. Interviews reveal widespread anxiety about the effects of AI on site traffic, audience behavior and news business models. Yet relatively few projects focus on revenue, subscriptions or fundraising.
Sixth, the barrier to building with AI has dropped, so journalists and small teams can develop prototypes and tools quickly. The cost of a product has fallen to the point where experimentation is no longer limited by technical capacity.
Overall, AI adoption across the organizations in this report is uneven, concentrated in large national outlets, well-funded digital-native organizations and some nonprofits. Three years into the generative AI era, most newsrooms are using AI to improve existing workflows, while a smaller group is experimenting with new ways of reporting and with audience facing products. The most prominent projects are those that apply AI to solve specific journalistic problems, not necessarily the ones that use the most advanced technology.
The entire dataset for this report was released publicly together with the study; readers can search, filter and explore all 160 projects and view profiles in the database. And read the full report here.