Perhaps no technology since the personal computer has transformed writing as radically as generative artificial intelligence, and the effect it has had on newsrooms is no exception. In fact, according to a recent study, a growing number of published newspaper articles show signs of being created by AI.

The study, conducted by researchers at Microsoft, Pangram Labs, the University of Maryland and UMass, used AI detection technology Pangram to analyze nearly 200,000 U.S. newspaper articles published by more than 1,500 news outlets. It found that AI use is “widespread,” appearing in approximately nine percent of published newspaper articles.

According to the study, 5.24 percent of the articles analyzed were determined to be AI-generated and an additional 3.98 percent were deemed “mixed,” containing text created by both people and machines.

newspapersAI use in newspapers by topic.

When it comes to what journalism topics were most likely to contain AI, the weather came in first by a significant margin (28 percent), followed by technology (16 percent) and health (12 percent). News topics that were least likely to feature AI included politics, environment, religion, crime and sports.

In addition to being “widespread,” the study also found the prevalence of AI in journalism to be “uneven.” According to the study, AI appeared far more frequently in smaller, local papers (9.3 percent) than it did in publications with circulations of more than 100,000 (1.7 percent). The study’s authors suggest this discrepancy could be a reflection of the fact that large, national newspapers typically "enforce stricter editorial constraints on automation than local papers."

The study also found that AI use is highest in newspaper articles published in the mid-Atlantic and southern U.S., while the Northeast "remains relatively unaffected."

However, the study also included nearly 45,000 opinion pieces from the Washington Post, Wall Street Journal and the New York Times and, interestingly, found that those op-eds were 6.4 times more likely to contain AI-generated text than news articles from the same publications. The study’s authors noted that many of the AI-flagged editorials included pieces “authored by prominent public figures.”

Finally, despite the widespread prevalence of AI in newspaper copy, the study also discovered that its use was rarely disclosed: A manual audit of 100 AI-flagged articles found that only five provided a citation that the content contained AI. This detail arguably strengthens the case for adopting a framework providing greater transparency in disclosing AI’s use in journalism.

“Overall, our audit highlights the immediate need for greater transparency and updated editorial standards regarding the use of AI in journalism to maintain public trust,” the study’s authors conclude.

A November survey by journalism nonprofit Trusting News found that 98 percent of respondents said they want to know anytime AI is used in news content, and an additional 48 percent said they’re uncomfortable with AI being used in the news, even if it’s “guided and verified before publication.”

The study, titled “AI Use in American Newspapers Is Widespread, Uneven, and Rarely Disclosed,” analyzed a dataset of 186,507 articles from the online editions of 1,528 publicly accessible newspapers between June and September 2025, in addition to 44,803 opinion articles published by the New York Times, Washington Post and the Wall Street Journal between August 2022 and September 2025.