Nearly one in 10 news articles published by American newspapers contains artificial intelligence-generated text, with smaller local outlets far more likely to use AI than major publications, according to new research examining 186,000 articles.
The study by University of Maryland researchers and AI transparency firm Pangram found 9.1 per cent of news articles across 1,500 US newspapers contained partial or complete AI-generated content, with usage rarely disclosed to readers.
Major newspapers with circulation exceeding 100,000 showed minimal AI adoption at 1.7 per cent of articles, whilst smaller local outlets recorded 9.3 per cent AI content — a disparity researchers linked to collapsing news economies and news deserts.
Geographic distribution revealed AI usage concentrated in the mid-Atlantic and southern states, with a correlation between communities lacking major newspapers and a higher frequency of AI-generated text. Corporate ownership patterns emerged, with Boone News Media papers containing 20.9 per cent AI content, followed by Advance Publications at 13.4 per cent.
“This disparity — that communities served by smaller papers and some corporate owners get more AI-made content than people in larger cities with bigger papers or different owner groups — is worrying, and may be a consequence of collapsing news economies, the result of news deserts,” said Bradley Emi, Pangram co-founder.
Opinions account for most AI
Opinion sections of prestigious publications showed higher AI adoption than news pages. The New York Times, Wall Street Journal and Washington Post opinion pages contained 4.5 per cent AI content versus 0.7 per cent in news sections—a 6.4-fold difference. Guest contributors accounted for most AI-generated opinion pieces rather than regular columnists, with researchers identifying 219 opinion articles containing AI content across the three publications.
The research employed Pangram’s AI detection technology to examine articles published this year. Most outlets failed to disclose their use of AI to readers, raising transparency concerns as newsrooms increasingly adopt automated content generation.
“Understanding the origins of our news is important, even vital, to being safe, informed, and able to make good decisions,” said Max Spero, Pangram co-founder and study co-author.
Pangram announced plans to launch an AI News Monitor that will provide regular updates on AI content trends in American journalism, aiming to track the technology’s role in news production as traditional business models collapse.
The findings arrive as local journalism faces existential pressures, with reduced staffing and revenue declines potentially driving smaller outlets toward automated content generation. Researchers warned the trend could create information inequality between communities served by well-resourced major newspapers and those dependent on struggling local outlets.
The study’s authors included Jenna Russell and Mohit Iyyer from the University of Maryland, Marzena Karpinska at Microsoft, and Destiny Akinode, Katherine Thai, Spero and Emi from Pangram.
