Newswise — New York City (October 23, 2025) – A new one research study released today reveals that more than 9% of all news published in U.S. newspapers contains at least some AI-created text.
The paper also reveals troubling trends related to the areas where AI-infused news content is most common, and also examines AI-produced content on the editorial pages of America’s most prestigious newspapers.
“We were able to look at 186,000 articles published by 1,500 journals this year and found that almost one in ten had at least some AI-created content,” said Max Spero, one of the journal’s authors and co-founder of the AI Transparency Society. Pangramwho provided the technology for the research. “The overall figure, nine percent, is surprising, but what we also discovered about where and perhaps why this is happening should be concerning,” he said.
The paper was co-authored by Jenna Russell and Mohit Iyyer of the University of Maryland, Marzena Karpinska of Microsoft, and Destiny Akinode, Katherine Thai, Max Spero and Bradley Emi of Pangram.
The article finds that “the use of AI in published articles is increasingly common but rarely disclosed. In our recent news dataset, 9.1% of articles are labeled by Pangram as AI-generated or mixed.” The paper adds: “Digging deeper, we see that AI use is unevenly distributed: it is much higher in small, local news outlets than in nationally distributed newspapers, and particularly concentrated in the mid-Atlantic and southern states of the United States. »
The paper found a correlation between communities lacking a major, widely-circulated newspaper and an increased frequency of AI-created text in news articles. In fact, the newspaper found that the use of AI in the news pages of major daily newspapers was quite limited: only 1.7% of articles in newspapers with circulations of more than 100,000 were partially or entirely generated by AI.
Examination of news articles created or containing AI also showed a correlation with news owners, with some major companies’ newspapers containing significant AI content. “Boone News Media has the highest percentage of partial or full AI content detected (20.9%), well above the next highest, Advance Publications (13.4%),” the paper said.
“This disparity – that communities served by small newspapers and some business owners get more AI-created content than residents of larger cities with larger newspapers or different ownership groups – is worrying and could be a consequence of the collapse of the information economy, a result of information deserts,” said Emi, co-founder of Pangram.
It is also troubling that in most cases the use of AI-created or infused information has not been disclosed to readers.
AI-generated content is not limited to news coverage, according to the report – surfacing AI texts on the opinion pages of three major newspapers – The New York Times, The Wall Street JournalAnd The Washington Post.
Although the use of AI on these pages was low, only 4.5% in total, it was still significantly more common than the use of AI on the news pages of these media outlets (0.7%). Most opinion pieces containing content on AI came from guest contributors rather than regular columnists, according to the newspaper. The research identified 219 opinion articles in these three newspapers containing content on AI.
“Opinion articles published in the New York Times, WaPo and WSJ are 6.4 times more likely to contain the use of AI than contemporary news articles from the same three newspapers,” the paper said.
“Understanding the origins of our news is important, even vital, to being safe, informed and able to make good decisions,” Spero said. “As the media and technology landscapes evolve rapidly and significantly, it is critical to stay attuned to news creation,” he said.
To help readers and news watchers find and track trends in AI usage across news and opinion pages, and to coincide with the release of the report, Pangram is also launching a “AI News Monitor», which will regularly update and publish the data from this first report.
“We are not alone on this,” Spero said. “Transparency in the use of AI is a core value for us, and we strive to provide people with good information about AI, especially when it comes to news and commentary,” he said.
About Pangram
Pangram Labs is the technology leader in AI sensing systems, outperforming other sensing vendors in accuracy, reliability and information transmission. Thousands of businesses rely on Pangram’s detection systems, primarily to evaluate and process public reviews of products and services, many of which are compromised by AI. Founded by Stanford University classmates, Pangram is gaining traction in the education market as a precision alternative for assessing the authenticity of student work.