NewsGuard and Pangram Labs Track AI Spam Sites

An Innovative System to Counter AI Content Farms

Media analysis company NewsGuard, in collaboration with AI detector Pangram Labs, has developed a real-time system to identify websites that massively publish content generated by artificial intelligence. This content, often riddled with factual errors, is produced by what are known as "AI content farms." Currently, 3,006 of these sites have been detected, a figure that has doubled in just one year. Each month, between 300 and 500 new sites emerge, reflecting a rapid and concerning growth.

The identification process relies on an automated AI detection software developed by Pangram Labs. This software first identifies sites where a significant portion of the content is generated by AI. Then, human analysts review the results to eliminate false positives, ensuring the accuracy of the detections.

Criteria for Identifying AI Content Farms

For a site to be labeled as an AI content farm, three conditions must be met. First, a significant portion of its content must be generated by AI. Second, the site must not disclose this fact. Finally, the site's layout must give the impression that the articles were written by human journalists. These sites often have generic names such as "Times Business News" or "Business Post" and publish dozens of articles per day.

Impact on Brands and the Spread of Propaganda

The lack of editorial oversight on these sites allows AI to generate completely fabricated stories. For example, in October 2025, the site "News 24" published a false report claiming that Coca-Cola threatened to withdraw its sponsorship of the Super Bowl if rapper Bad Bunny performed during the halftime show. NewsGuard pointed out that Coca-Cola is not even a sponsor of the Super Bowl. Nevertheless, ads from Expedia, AT&T, YouTube, Priceline, Hotels.com, Skechers, and GoDaddy were all displayed on this site.

Another site, "CitizenWatchReport," propagated the false claim that two U.S. senators had spent $814,000 on hotels in Ukraine. This story was picked up and amplified by state media in Russia targeting audiences in the United States.

NewsGuard indicates that the system also tracks sites managed by adversarial actors from Russia, China, and Iran. Among the identified AI content farms, 358 were linked to Storm-1516, a pro-Russian influence operation that publishes misleading content on sites designed to mimic local newspapers in the U.S. and Europe. NewsGuard noted in a previous study that the actual number of these farms is likely much higher, as current detection methods are not perfect.

Advertisers Trapped by AI Content Farms

Many reported sites are so-called "made-for-advertising" pages, specifically built to siphon off programmatic advertising dollars with low-quality content. NewsGuard previously reported that 141 well-known brands had run ads on this type of site over a two-month period.

The new data stream is designed to help advertisers avoid placing their ads on these sites. It integrates into buying platforms like The Trade Desk or can be licensed directly by brands and agencies.

AI content farms also appear in Google services like Google News and Google Discover, where they gain significant visibility. Google seemingly appears unable or unwilling to filter them out. Often, the company likely profits from ads displayed on these sites through its own AdSense program.