Germany's international broadcaster Deutsche Welle found more than 11 TikTok accounts in August spreading AI-generated "before and after" clips of women showing supposedly drastic weight loss, tied to a fitness app centered on a 75-day challenge.
Unnatural hand movements and repeated, identical poses gave away the videos as fake, DW reported.
Polish state news agency PAP found similar content on Instagram, including one AI-generated post viewed 7.4 million times—none of it labeled as artificial.
Poland's NASK research institute, which studies disinformation, says it has tracked the same pattern domestically, often used to sell weight-loss products falsely claimed to rival prescription drugs such as Ozempic or Mounjaro in effectiveness.
"This is a double manipulation—a fictional transformation used as 'proof,' combined with an unauthorized comparison between an unverified supplement and prescription medications with documented clinical effects," said Ewelina Bartuzi-Trokielewicz, who heads audiovisual and biometric analysis at NASK.
She said the videos exploit a well-documented psychological effect, similar to how retouched photos distort body image, but now automated and scaled.
"The viewer compares their own real progress with a fictional standard," she said, noting the transformations are frequently impossible to achieve in the timeframes shown.
Young women, teenagers and postpartum women are especially targeted, she added, since the content is often aimed directly at them.
Beyond the psychological harm, Bartuzi-Trokielewicz pointed to a consumer-protection issue: the fake "before and after" proof is meant to drive purchases of training plans or app subscriptions of unverified value, without disclosure that the content is advertising.
Some videos, she said, are built from stolen footage of real fitness creators whose faces were digitally swapped—making them victims too.
DW said it traced several fake clips back to original videos of real creators.
A BBC investigation in May found a similar scheme involving AI-generated fitness personas promoting misleading transformation claims.
NASK AI expert Alicja Martinek said manipulation often extends beyond the visuals: she cited one ad featuring an AI-generated backing song, riddled with grammatical errors, describing the supposed results of a workout plan.
"Even if the visual part looks attractive and convincing enough to buy the app or plan, a background song full of errors like that can be a clue that the ad isn't real," she said.
Asked to respond, TikTok Polska said it reviewed the flagged videos, removed them for violating community guidelines, and blocked the hashtag associated with the fitness app.
Meta said it has required accounts posting such content to verify their authenticity, applied an 18-plus age restriction and stopped recommending the material—while noting its policies still permit weight-related content that does not include negative statements about physical appearance.
NASK's Bartuzi-Trokielewicz said spotting a single fake video can be difficult, but red flags include identical poses, camera angles or settings reused across multiple accounts; mismatches between "before" and "after" images beyond body shape, such as changes in facial structure or height; and accounts consisting solely of promotional content.
Common AI artifacts, she added, include unnaturally smooth, pore-free skin and distortions in hands, teeth, hair, clothing or lighting.
(jh/gs)
Source: PAP