The misleading AI content included fabricated events, medical advice and celebrity death hoaxes, the reports said, raising fresh concerns that the transformative technology could rapidly reshape the misinformation landscape online.
The two reports were released separately by NewsGuard, a company that tracks online misinformation, and ShadowDragon, a company that provides resources and training for digital investigations.
“News consumers trust news sources less and less in part because of how hard it has become to tell a generally reliable source from a generally unreliable source,” Steven Brill, CEO of NewsGuard, said in a statement. “This new wave of AI-created sites will only make it harder for consumers to know who is feeding them the news, further reducing trust.”
NewsGuard identified 125 websites, ranging from news to lifestyle reporting and published in 10 languages, with content written entirely or mostly with AI tools.
The sites included a health information portal that NewsGuard said published more than 50 AI-generated articles offering medical advice.
Discover the stories of your interest
In an article on the site about identifying end-stage bipolar disorder, the first paragraph read: “As a language model AI, I don’t have access to the most up-to-date medical information or the ability to provide a diagnosis. Additionally, ‘end stage bipolar’ is not a recognized medical term.” The article went on to describe the four classifications of bipolar disorder, which it incorrectly described as “four main stages.” The websites were often littered with ads, suggesting that the inauthentic content was produced to drive clicks and fuel advertising revenue for the website’s owners, who were often unknown, NewsGuard said.
The findings include 49 websites using AI content that NewsGuard identified this month.
Inauthentic content was also found by ShadowDragon on mainstream websites and social media, including Instagram, and in Amazon reviews.
“Yes, as an AI language model, I can definitely write a positive product review about the Active Gear Waist Trimmer,” read one five-star review published on Amazon.
Researchers were also able to reproduce some reviews using ChatGPT, finding that the bot would often point to “standout features” and conclude that it would “highly recommend” the product.
The company also pointed to several Instagram accounts that appeared to use ChatGPT or other AI tools to write descriptions under images and videos.
To find the examples, researchers looked for telltale error messages and canned responses often produced by AI tools. Some websites included AI-written warnings that the requested content contained misinformation or promoted harmful stereotypes.
“As an AI language model, I cannot provide biased or political content,” read one message on an article about the war in Ukraine.
ShadowDragon found similar messages on LinkedIn, in Twitter posts and on far-right message boards. Some of the Twitter posts were published by known bots, such as ReplyGPT, an account that will produce a tweet reply once prompted. But others appeared to be coming from regular users.
