# AI reaches 82% of journalists as deeper newsroom adoption looks inevitable > Source: > Published: 2026-08-05 02:43:41+00:00 AI's deeper move into journalism now looks less like a question of whether than how far. Coding showed how quickly AI assistants can move from optional tools into routine professional workflows. Journalism is following that path: adoption is already broad, even if handing a system an entire first draft remains uncommon. [Muck Rack's 2026 State of Journalism report](https://muckrack.com/resources/research/state-of-journalism), released on March 19th, found that 82% of respondents used at least one AI tool, up from 77% a year earlier. ChatGPT reached 47%, transcription tools 40%, Gemini 22% and Claude 12%, according to [Muck Rack's release](https://www.globenewswire.com/news-release/2026/03/19/3259178/0/en/muck-rack-s-2026-state-of-journalism-report-finds-82-of-journalists-use-ai.html). The headline number measures any use of an AI tool. It does not show that 82% of journalists use AI to research stories, draft copy or generate material for publication. The inclusion of transcription software, which predates the current wave of generative chatbots, makes the figure a measure of broad workflow adoption rather than AI authorship. That distinction matters, but it does not undermine the direction of travel. AI took hold in coding by becoming useful inside an existing workflow, not by replacing the entire job at once. The journalism data shows the same initial pattern: tools are gaining ground first in bounded, repetitive tasks while humans retain responsibility for reporting, verification and final copy. Muck Rack's published materials also give different sample sizes. Its report page says the survey drew responses from nearly 1,100 journalists, while its release describes responses from nearly 900. The public materials do not reconcile the two figures, leaving unclear whether the smaller number represents respondents who completed a particular section. ### AI is spreading through the production chain A more granular [Reuters Institute survey](https://reutersinstitute.politics.ox.ac.uk/ai-adoption-uk-journalists-and-their-newsrooms-surveying-applications-approaches-and-attitudes), published in November 2025, shows where AI has entered the production chain. The broadly representative study of 1,004 UK journalists found that 56% used AI professionally at least weekly and another 27% used it less frequently. Language processing dominated. Forty-nine percent used AI for transcription or captioning at least monthly, 33% for translation and 30% for grammar checking or copy editing. Use fell to 22% for story research, 16% for brainstorming, 16% for generating parts of articles such as headlines and 10% for first drafts. The Reuters Institute findings also show that adoption inside organizations remains shallow. Sixty percent of respondents reported some newsroom integration, overwhelmingly describing it as limited. The same share said their outlet had established at least one AI protocol or guideline, while 32% said their newsroom provided AI training. An earlier [Associated Press-backed study](https://www.ap.org/the-definitive-source/products-and-services/ap-survey-reveals-ais-impact-on-newsrooms/) reached a similar top-line conclusion using a broader question. In late 2023, 73.8% of 290 respondents said they or their organization had used generative AI in some capacity. The sample leaned toward editors and executives: editors made up 34.5%, executives 20% and reporters 18.3%. Among 181 respondents who described actual applications, 69.6% mentioned text production. That category covered headlines, social posts, newsletters, outlines, summaries, rewriting and story drafts. The survey documented newsroom experimentation without establishing how much published reporting was written by AI. ### Leading newsrooms are formalizing assistant use Policies at large news organizations increasingly permit bounded assistance while keeping responsibility with journalists. That formalization is another sign that AI is becoming infrastructure rather than remaining an experiment. On July 23rd, 2026, [AP updated its newsroom standards](https://www.ap.org/the-definitive-source/announcements/ap-updates-newsroom-standards-for-artificial-intelligence/) to approve AI for early-stage research, document summaries, transcription, translation, headline suggestions, grammar and search optimization. AP requires journalists to review and edit the output and says AI cannot replace reporting, sourcing, editorial judgment or verification. [Reuters' journalistic standards](https://reutersagency.com/about/standards-values/) require independent verification of AI-generated facts, sources and claims. Reuters also holds individual journalists accountable for their work regardless of which tools assisted them. [The Washington Post's AI policy](https://www.washingtonpost.com/policies-and-standards/) allows uses such as translation and searching large collections of documents or images, with verification required because accuracy cannot be assumed. [The Guardian's March 2026 guidance](https://www.theguardian.com/help/insideguardian/2023/jun/16/the-guardians-approach-to-generative-ai) requires human oversight and senior-editor permission when significant generated elements appear in published work. Those rules put AI closest to the parts of journalism that can be accelerated without handing over final factual judgment: searching, sorting, transcribing, translating and reshaping material already gathered by reporters. They also give newsrooms a framework for expanding use as the tools improve. ### Published AI text remains uncommon at large papers A separate audit of what newspapers published provides a check on the survey data. University of Maryland researchers analyzed 186,507 articles from 1,528 US newspapers published between June 15th and September 15th, 2025. Their [preprint study](https://arxiv.org/abs/2510.18774), conducted with researchers affiliated with Pangram Labs, used Pangram's commercial detector to classify text as human-written, mixed or AI-generated. The detector flagged 9.1% of the full newspaper sample as mixed or AI-generated. The rate was 1.7% among papers with circulation above 100,000, according to the [University of Maryland's summary](https://www.umiacs.umd.edu/news-events/news/report-ai-use-newspapers-widespread-uneven-and-rarely-disclosed). Among news articles from The New York Times, The Washington Post and The Wall Street Journal, the study reported a 0.71% rate during the comparable period. Opinion pages produced a different result. The researchers found AI indicators in 4.56% of opinion articles from those three publications, with guest contributors accounting for much of the difference. The paper cautioned that a mixed classification cannot establish whether AI supplied substantive writing or made smaller edits. The surveys and publication audit capture different stages of the same shift. AI is already common newsroom infrastructure, with its strongest foothold in transcription, translation, research support, editing and production. First-draft generation remains a minority practice, especially inside large newsrooms with strict rules around sourcing and verification. Coding's rapid embrace of AI assistants supplies the clearest precedent for how quickly a professional workflow can change once utility is established. Journalism has already established that utility. The remaining questions are how deeply the tools spread, which tasks they absorb and what controls newsrooms impose along the way, not whether AI becomes a permanent part of the work.