Twitter Using AI Fact-Checking Technology

Last month, a misleading attempt to fact-check protest footage highlighted growing concerns about artificial intelligence’s role in combating misinformation on social media platforms.

The incident began when MSNBC shared a video on X (formerly Twitter) showing large crowds at a No Kings protest in Boston. A draft Community Note soon appeared claiming the footage actually dated from 2017—not October 2025. Though this note was never officially approved by X’s moderation system, screenshots of it spread rapidly across social media.

Senator Ted Cruz amplified the false claim before deleting his post. Fact-checkers at NewsGuard and the BBC later confirmed the footage genuinely depicted the recent No Kings demonstrations.

“This is the first time I can remember a Community Note that was screenshotted before it was even fully rated that spread that false claim,” said Sofia Rubinson, a senior editor at NewsGuard who investigated the incident.

What made this case particularly notable was the source of the erroneous fact-check: an AI bot called “Zesty Walnut Grackle,” one of X’s top Community Notes contributors.

The episode exemplifies the rapid shift away from professional fact-checking on major social platforms. When Elon Musk acquired Twitter in 2022, he laid off many staff responsible for content moderation and pivoted to the Community Notes program, which relies on users themselves to provide context or corrections to posts.

In September, X began allowing AI-written contributions to the program. A Tow Center analysis identified eight accounts utilizing X’s AI Note Writer API, collectively responsible for 5-10 percent of Community Notes visible to the public daily. These bots can be created by any user with verified credentials who isn’t already participating in the program.

Unlike traditional fact-checking, Community Notes require consensus from X users with diverse viewpoints. For a note to appear publicly, users who have disagreed on other notes in the past must vote to agree it’s helpful. Most notes—over three-quarters of those written since September, by both humans and AI—never achieve the necessary consensus to be displayed.

When they do work, Community Notes can effectively reduce misinformation spread. Research published in the Proceedings of the National Academy of Sciences found that tweets with visible notes received fewer retweets and likes. “The sooner a note is attached to a post pointing out the misinformation, the better the chances are of reducing further engagement with the post,” explained Martin Saveski, an assistant professor at the University of Washington who co-authored the study.

The Tow Center analysis revealed that AI-written notes aren’t inherently better or worse than human contributions. Users rated AI notes as helpful at similar rates to human-written ones, though quality varied significantly between different AI bots.

Researchers found numerous inaccurate AI-generated notes, including some that displayed confusion about basic facts, like incorrectly stating President Trump’s term had ended in 2021 or referring to him as a “private citizen” despite currently serving his second term. In these cases, human reviewers prevented the notes from being published.

The AI bot responsible for the No Kings misinformation eventually corrected itself, acknowledging its error with a screenshot of its incorrect note and confirming the footage was indeed from 2025.

X’s shift toward user-based fact-checking appears to be influencing other platforms. Meta, which abandoned in-house fact-checking earlier this year, has developed a similar community notes program.

“People who are more skeptical of fact-checkers tend to thrive on X and tend to give a little bit more weight to these Community Notes as pretty reliable,” Rubinson noted.

As AI becomes increasingly integrated into fact-checking systems across the internet, the critical question remains whether it can perform this role with sufficient accuracy to be truly effective in combating misinformation—particularly during politically sensitive periods when accurate information is most crucial.