“Five strict beauty rules they never showed you on camera, but every woman on Love Island had to follow, from someone who worked behind the scenes for nine years,” a woman wearing a baseball hat and gold hoop earrings says teasingly to the camera.
Her “insider” tips are eyebrow-raising. Contestants were, she claims, required to eat a concoction of papaya and cucumbers to drop 11 pounds in one week. They drank lemon balm tea that she calls “Nature’s Ozempic” at the request of producers to help “flatten their stomachs,” she says. And finally, they apparently used one product before any filming days: an exfoliator by the brand, Kollekt.
If you’re mindlessly scrolling through your FYP, you might not notice that this isn’t a real insider, nor are these real tips. In fact, it's not even a real person, but a realistic synthetic avatar named Tonya. Available for free, public use on HeyGen, an AI video generator tool headquartered in Los Angeles, Tonya is one of hundreds of avatars on the platform that anyone can turn into a video like this Love Island one.
The video, posted to TikTok, amassed more than 1.1 million views before it was removed — but a plethora of similar videos litter Instagram and TikTok right now.
A quick search using the keywords “love island beauty secrets” will yield dozens of results for synthetic avatars reciting a usual script, typically starting with a line like, “here’s the secrets they don’t show on camera” or “I worked behind the scenes for seven years.” AI-generated content has been used to sell wellness and beauty products before, but this recent iteration of videos suggests that synthetic avatars are being employed in increasingly culturally literate ways.
“Whatever will maximize engagement on the platform goes and wins out,” said Ben Winters, the Director of AI at the Consumer Federation of America. “[It] truly doesn't matter to them how people are impacted [...] so you see younger women actually being targeted.”
In more dangerous situations, deepfakes have been used for financial scams or to impersonate politicians during elections to disrupt democratic processes. Recently, synthetic avatars are showing up on your FYP as clever advertising tools that know what hot topic has captured people’s attention spans — and spewing pseudoscientific wellness claims in the process. The videos are posted under social media accounts, typically with some combination of buzzy phrases like “health secrets,” “beauty tips” or “healthy girls.” It’s not clear who, exactly, is creating the videos, but they often seem to be trying to sell a specific product.
TikTok content
One video with a synthetic avatar named Linda, also generated from HeyGen, follows a typical script: “The dark side of Love Island nobody is allowed to talk about. I auditioned for season 7 and the things they made us do to look good on camera are absolutely insane.” In the video, which has 28,000 likes and over 700,000 views, the avatar talks about rubbing castor oil to make stretch marks disappear and drinking lemon balm tea for a thin waistline. At the end of the clip, the avatar gushes over a lemon balm extract from beauty brand, Eliré. Another video, using an avatar from the AI video generating app Captions, also promotes the lemon balm extract and talks about how contestants had to “slam apple cider vinegar shots mixed with cayenne.”
Another avatar, also generated by Captions, claims to have worked on the show for seven years and alleges that contestants drank a blend of chia seeds and apple cider vinegar “to help them sh*t fast and make their stomachs look flat.” In the end, the avatar states that all contestants had to use dark spot remover pads from JiYu — a Korean beauty brand that has been featured in Vogue. Around the same time that season six contestants JaNa Craig and Kenny Rodriguez announced their split, the same avatar was used in a video talking about the breakup — and also promoting the toner pads.
TikTok content
On Instagram, the situation is not all that different. One video with an avatar from Captions (that has since been removed from the platform) said contestants had to drink spearmint tea to help offset “hormonal imbalances” and used filtered shower heads from the brand Afina. There is a surplus of these videos online, spanning from avatars promoting “de-bloating supplements” to unproven weight loss tips, all under the guise of entertainment gossip. Are you feeling the uncanny valley of it all yet?
Teen Vogue has reached out to each of the brands mentioned in this article but has not yet heard back.
In an untrustworthy information ecosystem, AI abuse and scams thrive, according to Winters. In some ways, the information the avatars are being used to disseminate is not that different from early 2000s tabloids publishing harmful diet tips — the technology is just more targeted and pervasive.
“It is marketing, but lots of marketing is bad and manipulative and leads to harm,” Winters said.
Earlier this year, an iProov study that surveyed 2,000 US and UK consumers found that only 0.1% of participants could accurately distinguish deepfakes from real content in all videos and images they were shown.
“A few months ago, I could give you visual tips that often worked: watch for unnatural blinking, lip sync that’s slightly off or a strange-looking background,” said Tina Nikoukhah, the research director of GetReal, a digital security startup. “Those can still help sometimes, but more and more videos no longer show these obvious flaws.”
That’s why videos like these Love Island ones are so successful. Many users likely can’t detect that they’re AI driven, instead believing that insiders are giving them real diet tips to look like their favorite Islanders. It taps into both the cultural obsession with thinness pervasive among young women especially, and the many ways AI is infiltrating our cultural consciousness.
Beyond the visual, Nikoukhah said it’s important to check the context. Nowadays, you might find some luck with a reverse image search to find the original source of a video, which could help you find other existing videos with the same AI avatars — although, the free, open-source tools are not always the most accurate.
But that only works if the person in the video isn’t familiar to you. AI can also be used to generate avatars of real people, like celebrities, that can make it more difficult to detect when something is fake. Several celebrities have been deepfaked before. Last year, a deepfake of Taylor Swift promoting Le Creuset cookware deceived some fans. In July, a viral video of season seven contestant, Huda Mustafa, promoting a fragrance sold on TikTok Shop, Shein, and Amazon fooled several people who bought and posted reviews of the scent.
“Synthetic avatars make it easy to create endless ‘perfect’ spokespeople who never age, work 24/7 and can be tuned to match any trend,” Nikoukhah said. “The same techniques could be used to impersonate a CEO to push a fake business deal, or fuel financial fraud, romance scams, and other schemes.”
While the speed at which AI is developing makes it more difficult to distinguish real from fake, one thing is certain: don’t necessarily expect social media platforms to explicitly disclaim whether the content you’re seeing is AI-generated — they are not legally mandated to do so, according to a spokesperson for Reality Defender, a deepfake identification company. “That is the bare minimum that social media platforms should be performing,” the spokesperson said. “And they're not doing that.” While TikTok and some other platforms require creators to label all AI-generated content, not all of the synthetic avatars identified by Teen Vogue had a disclosure.
“My suggestion is [to] speak with your representatives and tell them that some level of AI legislation that requires social media platforms to be proactive and not reactive is super important,” the spokesperson said.
In July, the White House released “America’s AI Action Plan,” which introduced provisions to remove regulatory barriers on AI and called on the Federal Trade Commission to review investigations from previous administrations to ensure they do not “unduly burden AI innovation.”
“Right now the consumer is given the worst possible deal online,” Winters said. While most states do have laws enacted to regulate AI-generated intimate deepfakes, the U.S. doesn’t currently have comprehensive federal legislation regulating AI development or restricting its use. “You have very limited privacy protections in most states at least, and you have almost no regulation of AI at the state level or at the federal level.”
Consumer Reports, a consumer advocacy non-profit, responded to the Trump Administration’s plan, stating that it could lead to AI developers having “free rein” to create harmful products, like therapy chatbots, voice cloning and explicit deepfake generators.
“It is a little bit of a blank check,” Winters said. “[It’s] like, ‘do whatever the hell you want’ to AI companies. There's no incentive to try to be responsible.”
