AI use in American newspapers is widespread, uneven, and rarely disclosed
A 2025 audit of 186,000 articles from 1,500 American newspapers reveals that approximately 9% contain AI-generated content, with usage disproportionately concentrated in smaller local outlets, specific topics, and opinion pieces, yet remains largely undisclosed despite its prevalence.
Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the American newspaper industry as a massive, bustling kitchen with 1,500 different restaurants, ranging from tiny, family-run diners to huge, famous five-star chains. A team of researchers recently decided to taste-test the food being served in this kitchen to see how much of it was actually cooked by the chefs (humans) versus how much was prepared by a high-tech robot (AI).
Here is what they found, explained simply:
1. The "Robot Chef" is Already in the Kitchen
The researchers looked at 186,000 articles published in the summer of 2025. They used a very sharp "taste-tester" tool called Pangram (think of it as a super-smart food inspector) to check if the writing was human or machine-made.
- The Result: About 9 out of every 100 articles had some help from a robot. Some were fully cooked by the robot, while others were a "mixed" dish where a human chef added the final seasoning to a robot's base recipe.
- The Surprise: The robot isn't just in the fancy restaurants. It's actually much more common in the small, local diners (local newspapers) than in the big, famous chains (national papers). It's also showing up more in specific "menu items" like weather reports, science, and technology, and less in serious topics like war or crime.
2. The "Opinion Section" is a Robot Hotspot
The researchers also looked at the "Opinion" pages of the three biggest newspapers in America: The New York Times, The Washington Post, and The Wall Street Journal.
- The Result: Opinion pieces were 6.4 times more likely to have robot help than regular news stories.
- Who is using it? It's not just the staff writers. Many of these robot-assisted opinions are written by guest stars—famous people like Nobel Prize winners, Senators, Governors, and CEOs.
- The Trend: In 2022, before the big AI boom, almost no one used it for opinions. By 2025, that number jumped up significantly.
3. The "No-Label" Problem
Here is the biggest issue: Nobody is telling you.
- The Secret: The researchers manually checked 200 articles that the robot-inspector flagged as AI-written.
- 96.5% of the authors didn't say a word about using AI.
- 94% of the newspapers didn't put up a sign saying, "Hey, a robot helped write this."
- The Analogy: Imagine going to a restaurant, ordering a steak, and the chef hands you a plate. You don't know if it's a real steak or a high-tech meat substitute. The restaurant doesn't tell you. You just eat it. That is what is happening with these news articles.
4. The "Fake Ingredients" Risk
When the researchers looked closely at the robot-written articles, they found a lot of hallucinations. In the world of AI, a hallucination is when the robot speaks confidently about something that simply isn't true.
- The Stats: Articles flagged as AI-generated were 8.2 times more likely to contain fake facts (like wrong dates, made-up quotes, or incorrect statistics) compared to human-written news.
- The Danger: Because these articles often don't say they were written by a robot, and because they contain more errors, readers might believe false information without realizing it.
5. Why the Robot is Everywhere in Small Towns
The study found that the robot is most popular in small towns and in languages other than English.
- Small Towns: Small newspapers often have very few staff members. They might be using the robot to help write weather reports or local sports scores because they are short on time and money.
- Other Languages: Articles in Spanish, Portuguese, and Vietnamese had a much higher rate of robot help (31%) compared to English articles (8%). This suggests that translation tools or automated writing are being used heavily to serve these communities.
The Bottom Line
The paper concludes that AI is already a major part of American journalism, but it is being used unevenly and secretly.
- It's not all bad: Using a robot to check grammar or write a simple weather report is like using a calculator; it can be helpful.
- The problem: When a robot writes a complex story or an opinion piece and no one tells you, it breaks the trust between the newspaper and the reader. It's like a magician hiding their tricks; once the audience finds out the tricks were done by a machine without being told, they stop trusting the show.
The researchers are asking newspapers to start putting up "Robot-Assisted" signs on their menus so readers know exactly what they are consuming.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.