AI use in American newspapers is widespread, uneven, and rarely disclosed

TL;DR

This paper quantifies AI-generated content in the U.S. press by auditing a large cross-sectional corpus of about 251k articles from mid-2025 using the Pangram AI detector. It analyzes three datasets (recent_news, opinions, ai_reporters) with topic, language, ownership, and circulation metadata, revealing that roughly $9\%$ of articles are AI-generated or Mixed, with higher prevalence in local outlets, weather and science/tech topics, and non-English languages. The findings show that AI use is substantially more common in opinion pieces at top newspapers (about $6.4\times$ more likely than news articles), yet disclosure remains rare, with only a small fraction of AI-flagged items explicitly labeling AI involvement. The work also introduces public datasets, analysis code, and an interactive dashboard, and argues for editorial policies to improve transparency and reader trust as AI-assisted journalism expands.

Abstract

AI is rapidly transforming journalism, but the extent of its use in published newspaper articles remains unclear. We address this gap by auditing a large-scale dataset of 186K articles from online editions of 1.5K American newspapers published in the summer of 2025. Using Pangram, a state-of-the-art AI detector, we discover that approximately 9% of newly-published articles are either partially or fully AI-generated. This AI use is unevenly distributed, appearing more frequently in smaller, local outlets, in specific topics such as weather and technology, and within certain ownership groups. We also analyze 45K opinion pieces from Washington Post, New York Times, and Wall Street Journal, finding that they are 6.4 times more likely to contain AI-generated content than news articles from the same publications, with many AI-flagged op-eds authored by prominent public figures. Despite this prevalence, we find that AI use is rarely disclosed: a manual audit of 100 AI-flagged articles found only five disclosures of AI use. 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.

Figures (25)

• Figure 1: Map of the United States showing the AI use (%), number of articles, and number of unique newspapers for each state. States in the mid-Atlantic and southern US exhibit markedly higher AI use than other states. Note that this plot only considers articles written in English (see \ref{['fig:non_english_state_heatmap']} for others).
• Figure 2: AI use by topic in recent_news. Weather, science/technology, and health exhibit higher AI use than topics like conflict, crime/justice, and religion.
• Figure 3: Heatmap of AI use by publication owner and article topic in recent_news. Some owners disclose AI use for specific content, such as https://www.advance.com/ for weather reports, but the biggest adopters, such as https://boonenewsmedia.com/ |, use AI broadly across many topics. Note that only 52% of recent_news has ownership information, and only topics that have at least 5 articles per owner are visualized in the heatmap.
• Figure 4: AI use in recent_news is more frequent in languages other than English. The most prominent such languages include Spanish, Portuguese, Vietnamese, French, and Polish.
• Figure 5: An example of an article found in print which was identified as AI-Generated in recent_news. Even readers who get their news exclusively from print media are exposed to AI-generated articles. https://www.pangram.com/history/13fec041-70ef-4cee-ab57-00b9c7541c04/https://www.washingtoninformer.com/trump-border-wall-controversy-growth/