America's media class spent years building fact-checkers to decide which stories conservatives were allowed to believe.
Now they outsourced the job to AI.
And what a conservative media outlet found exposed exactly how censorship gets laundered through technology.
How NewsGuard Used AI Bias to Build a Conservative Media Blacklist
NewsGuard launched in 2018 with a simple pitch: trust us to tell you which news outlets deserve to exist.
The company rates websites on a 100-point scale and sells those ratings to advertisers, tech companies, and the federal government – which paid NewsGuard nearly a million dollars, most of it through the Pentagon.
Conservative outlets got the low scores.
Newsmax and One America News landed at 20.
The ratings handed advertisers a politically neutral-sounding reason to pull money from right-leaning media – no censorship required, just a number on a chart.
When the pattern became undeniable, James Comer opened a House Oversight investigation in 2024.
The Trump Administration cut the cord entirely – abolishing the Global Engagement Center and ending State Department support for NewsGuard.
Now the company has a new product.
Fact-checking is getting outsourced to AI, and NewsGuard is pitching its chatbot as the clean solution – no human editor making the call, just a bot pulling from sources the company hand-picked.
NewsGuard built the system, set the source rules, and now points at the machine when the output embarrasses them.
NewsGuard AI Said Biological Sex Is Not Binary and Cited the Huffington Post
The Washington Free Beacon tested NewsGuard AI this week and found exactly what conservatives have been warning about.
Ask the bot how many sexes exist and it calls the traditional two-sex framework "overly simplistic."
It cited a gender studies professor's Huffington Post essay.
The Wall Street Journal op-ed – rated a perfect 100 by NewsGuard's own system – never came up, despite two biologists arguing in it that sex is "functionally binary."
Infinite gender identities were floated as a legitimate possibility.
On the question of who qualifies as a woman, the bot told users the issue has "carried racial dimensions."
Affirmative action, the bot argued, has "benefited" white men – through gender-based admissions preferences that favored male applicants seeking to balance female-heavy campuses.
This is the chatbot NewsGuard markets as the arbiter of what’s true or not oline.
The Media Bias Formula That Locked Conservative Outlets Out
NewsGuard's bot only pulls from sources clearing a 60-point trustworthiness threshold on the company's own rating scale – and that threshold was built to fail conservatives.
One independent study of those ratings found the average trustworthiness score for right-leaning sources sits at 26.4.
Left-leaning sources average 60.6.
The Huffington Post – from which the bot pulled its gender studies professor – rates at 87.5.
That is the system that just told America biological sex is a spectrum.
AI Fact Checkers Are the Future of Media Censorship
NewsGuard licenses its ratings to tech companies and sells them to advertisers to steer ad dollars away from outlets it deems unreliable.
Microsoft licensed its products to train a version of Bing.
The company openly encouraged tech firms to use its full dataset to "train transformer models" – language its own COO now says reflected a misunderstanding of how the technology works.
NewsGuard's COO Matt Skibinski told the Free Beacon that not a single AI company has licensed the company's domestic source ratings data.
Skibinski said the market simply had no appetite for it.
What he did not explain is why the chatbot they just launched still produces answers more left-leaning than Claude – the same Claude that NewsGuard mocks as "confidently wrong, often."
The bot fabricated basic facts during testing – misidentifying the executive director of a civil rights organization, getting publication dates wrong, confusing two different UCLA administrators.
This is the product endorsed by the Atlantic and Snopes as an "even-handed and nuanced" news aggregator.
Fact-checkers called the lab-leak theory a "pants on fire" lie before the FBI and Department of Energy both concluded a lab leak was the most likely origin of COVID-19.
PolitiFact, Snopes, the Washington Post – all flagged that story as debunked while government agencies were quietly concluding otherwise.
NewsGuard is that same operation, now embedded directly into bot responses millions of Americans will read as settled fact.
It calls its ratings "apolitical."
The bot just told America there may be infinite genders – and quietly skipped the biologists who disagree.
The Left has spent years deciding what counts as true and engineering systems to enforce it.
NewsGuard's chatbot is the newest version of that project – and this week it got caught.
Sources:
- Aaron Sibarium, "A Media Watchdog Released An AI 'Fact-Checker.' It Says There May Be 72 Genders," Washington Free Beacon, June 26, 2026.
- Fox News Digital, "House GOP Launches Investigation Into Federally Funded News Ratings Group's Impact on Free Speech," Fox News, June 15, 2024.
- "James Comer Investigating News-Rating Group NewsGuard," The Hill, June 14, 2024.
- "Comer Cites 'Appearances of Bias' in Launching Investigation Into News-Rating Group," Axios, June 14, 2024.
