Researchers say a previously unknown network of hundreds of accounts on the X platform is using artificial intelligence to automatically respond to conservatives with positive messages about members of the Trump administration.
But as MAGA movement is divided over the administration’s handling of cases related to deceased sex offender Jeffrey Epstein, posts from these accounts have been revealed, containing conflicting statements about the case, revealing their LLM nature.
This network, tracked by social media analytics firm Alethea and researchers at Clemson University for NBC News, includes more than 400 identified bot accounts, although the number could be much higher, according to the researchers. These accounts regularly praise key figures in the Trump administration, including Health and Human Services Secretary Robert F. Kennedy Jr. and White House Press Secretary Carolyn Leavitt.
As is often the case with bot accounts, the accounts monitored by NBC News have only a few dozen followers, and their posts are rarely seen. But a large audience doesn’t appear to be the goal. Their effectiveness, if anything, lies in the hope that they help create a partisan echo chamber and, collectively, can “sway public opinion,” said Darren Linvill, director of Clemson University’s Center for Media Forensics, which studies online disinformation campaigns.
“Their goal is not to generate engagement, but simply to occasionally appear in replies,” Linvill told NBC News.
The researchers declined to disclose details of how they identified the accounts, but noted that they shared several distinct trends. All appeared to have been created, apparently in batches, on approximately three specific days last year. These accounts frequently added punctuation to their posts, often unrelated to the conversation. They post almost exclusively in response to other users, often paying a fee for fact-checking, repeating similar opinions repeatedly, at short intervals. Sometimes, they respond to a post by repeating it verbatim.
It is unclear who is behind this network, nor which of the many AI-powered chatbots, widely available to the public, was used to operate it.
These bots have shown support for conservative figures since 2024, including Trump and other Republicans during the primaries, and have subsequently expressed enthusiasm for Trump’s presidency. While they have sometimes expressed conflicting messages—some have expressed sympathy for MSNBC anchor Rachel Maddow, for example—their messages have consistently supported MAGA figures until the recent Epstein controversy.
A significant portion of Trump’s supporters voted for him, convinced that Trump, a former friend of Epstein, would unveil a list of allegedly wealthy and powerful clients and bring justice to Epstein’s victims.
The accounts’ posts have only been divided since the beginning of the month, when Attorney General Pam Bondi announced she would no longer release any Epstein files, with some accounts displaying opinions that are almost the opposite of those of their users.
For example, at the same time last Saturday, one account on the network warned a MAGA supporter against judging Bondi too harshly, and another told Bondi, FBI Director Kash Patel, and Deputy Director Dan Bongino that they should resign because of the scandal.
When Bondi initially stated she would not release any additional files, another said she was “proclaiming her innocence,” as the Justice Department confirmed it had found no list of Epstein’s clients, reaffirming his suicide. Since then, the department has urged many people on X to revolt against the Trump administration.
“Retweet if you think Trump and his accomplices are lying to the public and calling us idiots. We will not be fooled by their shenanigans,” the post published Friday read.
While the motives of these accounts remain unknown, they appear to have been modeled after genuine MAGA social media accounts, whose messages tend to be more uniform, according to C. Sean Ip, head of investigations at Alethea.
“This divergence in reactions reflects the natural reaction of supporters of a second Trump administration,” Ip told NBC News. The behavior of these automated accounts is likely influenced by content posted by prominent influencers, and this shift reflects a general shift in attitudes among many Trump supporters.
For years, social media platforms have been rife with fake accounts designed to influence public opinion, whether by marketing agencies promoting products or by foreign governments and domestic groups disseminating political propaganda. But this situation has intensified with the rapid spread of intelligent chatbots capable of writing and posting persuasive messages without human intervention, and has been exacerbated by the reduction in content oversight by social networks.
The White House declined to comment, and the Department of Health and Human Services and X did not respond to requests for comment.
Researchers believe X is likely rife with fake accounts, although their extent is impossible to determine. Last year, researchers discovered a separate network of pro-Trump accounts powered by artificial intelligence on X. However, it’s impossible to understand the extent of their presence on the network. X disbanded much of its Trust and Safety team when billionaire Elon Musk bought the site in 2022, and since then, it has made it much harder for researchers to access the data.
