As artificial intelligence companies grapple with how to combat false claims and improve the reliability of chatbot responses, one media ratings organization is seeking a larger role in shaping how AI systems evaluate news sources.
NewsGuard, a company that rates news outlets on a 100-point credibility scale, has spent years marketing its assessments to advertisers seeking to avoid placing ads on websites it considers unreliable. More recently, the company has positioned its ratings and data products as tools that could help AI developers identify trustworthy information and reduce the spread of unverified assertions.
In a 2023 announcement, NewsGuard said its ratings could help AI models distinguish between credible and unreliable sources during the development process, warning that large language models risk becoming major distributors of inaccuracies without safeguards.
“Without tools like NewsGuard for AI,” the company said at the time, “AI language models risk becoming the next great misinformation superspreader.”
The effort has sparked criticism from free speech advocates and conservative groups who argue that NewsGuard’s ratings reflect subjective judgments that could introduce ideological bias into AI systems used by millions of people.
One of the most frequently cited examples involves NewsGuard’s ratings of certain conservative American media outlets compared to Chinese state-run media organizations.
According to NewsGuard’s publicly available ratings, China Daily, which is owned by the Chinese Communist Party’s Central Propaganda Department, receives a score of 44.5 out of 100. Several right-leaning American outlets score lower, including Newsmax at 20, The Federalist at 17.5, and One America News Network at 22.5.
NewsGuard argues those scores reflect its assessment of each outlet’s accuracy record rather than political orientation. The company says China Daily receives significant deductions for transparency and editorial independence concerns, but does not fail certain accuracy-related criteria because NewsGuard determined it did not publish enough demonstrably false claims to trigger additional penalties.
“Who decides what’s reliable and what’s not?” asked Gene Hamilton, president of America First Legal, which has criticized NewsGuard’s influence on content moderation policies.
The debate highlights broader disagreements over how AI systems should determine what information is trustworthy.
NewsGuard co-founder Steven Brill has long defended the company’s methodology, insisting that ratings are based on journalistic standards rather than politics. The company points to highly rated conservative outlets such as Fox News and the New York Post as evidence that political ideology is not the determining factor.
Still, critics note that NewsGuard’s leadership and advisory ranks have historically skewed left politically. Analyses of political donations by company executives and advisers have found significantly more contributions to Democratic candidates than Republican candidates.
The company also developed relationships with government agencies and technology firms during the Biden administration’s broader online censorship efforts. In 2021, NewsGuard received a Pentagon contract, and later launched services that aided technology companies target election-related dissent.
Questions about NewsGuard’s role in AI have intensified as chatbots increasingly replace traditional search engines and become a primary source of information for many users.
Mike Benz, founder of the Foundation for Freedom Online and a frequent critic of content moderation initiatives, argues that allowing ratings organizations to influence AI systems risks embedding contested political and editorial judgments into widely used technologies.
“The fear every citizen should rightly have is that a highly biased private firm could influence or effectively control the sources from which AI derives its decisions,” Benz said.
Supporters of credibility-rating systems argue that AI models need mechanisms to distinguish between accurate reporting and false claims. Critics counter that any effort to create centralized arbiters of truth risks importing human biases into systems that many users increasingly view as objective.
READ NEXT: LGBT Organization Urges AI Companies To Suppress Unfavorable Information
