The American soldiers who appeared in a TikTok video last month are handsome, muscular and, inexplicably, shirtless.
“We’re moving out shortly,” one says, sitting inside what appears to be a troop transport aircraft. “It’s going to be a long one. If you can, just send a little love our way.”
The video is a concoction, generated by artificial intelligence with the goal of enticing viewers to click. It is one of thousands, according to researchers who track them, that have deluged social media with fake images made to look like American service members on duty — at a time when real soldiers have been serving in harm’s way largely out of public view.
The source and the intent are not always obvious, the researchers said. Some, though not all, have been part of foreign influence operations, according to researchers, including accounts that during the war with Iran have circulated propaganda purporting to show American soldiers crying or quaking with fear.
One coordinated network of accounts on TikTok has digital traces to China, according to researchers affiliated with the University of Southern California who discovered it, though the exact purpose is not clear.
All of them exploit the support that most Americans share for military service, even when the wars themselves are unpopular, as is the one the United States and Israel started with Iran in February. Many draw thousands of likes and comments.
“There is a natural audience for this because Americans care about their military and the war is salient,” said Peter D. Feaver, a political science professor at Duke University and an expert on civil-military relations. He noted that expressing that support was “a genuine American ritual of our civic religion.”
Other similar accounts, based in places like Pakistan and Indonesia, appear to be motivated primarily by engagement farming, the practice of building large followings of Americans on accounts that can later be sold and repurposed.
Impersonating American service members has long been a tactic in online scams, but advances in A.I. tools have created new ways to deceive users.
One account on Facebook posted numerous videos of a young woman dressed in Army combat camouflage, identifying her as a Vietnamese American soldier, according to TruthScan, a company that tracks the use of A.I. in videos. The comments section includes a link to an account on OnlyFans, the platform known for adult sexual entertainment.
“It would be foolish to assume that there aren’t some nefarious people doing things with these accounts,” said Devan Leos, chief communications officer for TruthScan.
The company linked that account to a coordinated campaign this spring that also posted pro-Iran videos and, more recently, A.I. videos of Israeli soldiers. The accounts seemed to be operated by a single user in Pakistan, who did not respond to a request for comment.
Major platforms like Facebook and Instagram, both owned by Meta, as well as TikTok have policies against coordinated inauthentic behavior and rules requiring labels on videos generated with A.I. tools, but enforcement is inconsistent. And platform algorithms frequently promote them in user feeds.
During the height of the fighting in Iran, thousands of videos showed American soldiers — and some Israelis — visibly afraid, distressed or crying. According to a report at the time by Alethea, an online threat analysis company, a network of pro-Iranian accounts on X posted or shared the videos to suggest that the American troops involved did not support the war effort.
“I miss my mom and dad,” a woman in uniform said in one fake video that circulated, her voice cracking and buffeted by wind. “Follow my page. It gives me support.”
Iranian-linked accounts posted it on X with an identical message in Arabic, indicating the target audience was Iran’s Arab neighbors. “#MoralCollapse is evident on the faces and tone of the American soldiers,” the message said.
Researchers with the University of Southern California’s Information Sciences Institute have since documented a new coordinated network of more than 100 accounts on TikTok that has been regularly posting A.I. videos of American men and women in uniform making similar appeals for likes or follows.
Unlike soldiers in the Iranian-linked videos that appeared when the war began, those in the new ones do not seem distressed but are instead asking for recognition for the sacrifices they make for the mission.
Luca Luceri, a researcher with the institute who published an analysis of the network for a new company, DotNex, said the network’s campaign had generated nearly three million views and 444,000 engagements by users who liked or commented. Advances in A.I. not only make the videos more persuasive for some people but also make campaigns easier to organize at scale.
“The novelty isn’t the slop itself,” Mr. Luceri said. “What’s new is that this is fully automated, foreign-operated and targeting military families specifically through emotional content that real people responded to sympathetically.”
Digital traces in the accounts suggested links to China and Indonesia, he said, but without a clear connection to government influence operations.
The New York Times found numerous others on TikTok, Facebook and Instagram. TikTok and Meta did not respond to requests for comments on the proliferation of the videos, many of which have flaws that are hallmarks of A.I.
Details of uniforms and military equipment are inaccurate. Names are garbled on name tags. Many of the characters do things no soldier would be allowed to do, like boarding an aircraft shirtless or driving a Humvee in a combat zone without a helmet.
And yet they can be persuasive. On a post with one fabricated video, a Facebook user who identified himself as a Vietnam War veteran commented earnestly that the woman shown was not wearing her hair in keeping with U.S. Army regulations.
None of several accounts contacted through the platforms responded to requests for an interview.
Ellen Gustafson, co-executive director of We the Veterans and Military Families, a nonprofit, said the proliferation of generative A.I. tools had increased the scale of scams, cyberattacks and disinformation campaigns. She said that when alerted, the platforms often took down A.I. accounts and content, but that new ones soon emerged.
“It’s Whac-a-Mole,” she said. “I mean, they come right back up in another form.”
In addition to the threat of influence operations, she said, A.I. slop creates a distorted, often cartoonish view of wars and the people who serve in them, one she compared to video games depicting combat.
“Making real, noble and humble service into this gamified, bravado imagery is not cool,” she said. “It doesn’t necessarily invite the next generation of people to serve, because it paints a picture of only a certain type of person and a certain kind of military job, which is not realistic and is not what we need.”
