Fake AI-generated influencer goes viral during Bama Rush on TikTok

It starts with a shy smile and a nervous wave. A curly-haired redhead named Janie sits in her dorm room, hands clasped, introducing herself to the world on the first day of sorority rush at the University of Alabama. She’s wearing a pleated skirt from Abercrombie, a knit top from J. Crew, socks from Target, and sneakers from Nike. She lists off each item with the earnest, slightly rehearsed energy of a college freshman trying to make a good first impression. The video, an “outfit of the day” or OOTD post, quickly became a hit, amassing over 120,000 views and drawing a flood of adoring comments. People called her adorable, wished her luck, and followed along as she documented the whirlwind of rush week—the outfits, the nerves, the excitement, the chaos. She seemed like just another eager potential new member caught up in the viral TikTok phenomenon known as RushTok, where thousands of young women document their journeys into sorority life every year. But here’s the twist: Janie isn’t real. She doesn’t go to Alabama. She didn’t rush a sorority. She didn’t even get her clothes from J. Crew or Nike. Every detail of her videos—the dorm room, the outfits, the voice, even Janie herself—was generated by artificial intelligence. And her creator, a venture capitalist named Olivia Moore, made her to test just how easily the internet would fall for a beautifully rendered digital illusion.

Moore is an investment partner at Andreessen Horowitz, where she focuses on consumer AI companies. She has long been fascinated by how people create, interpret, and emotionally respond to AI-generated media. With image, video, and audio models becoming eerily human-like at breakneck speed, she believes we are about to cross the uncanny valley completely. “The image, video, audio models are getting very human-like, very fast. I think we probably cross the uncanny valley in the next six months,” she said in an interview. If that happens, she argues, the ability to tell real from fake—and our willingness to care either way—will define how we navigate a world flooded with synthetic content. So she created Janie as a social experiment, spending only $100 in credits to produce an entire week’s worth of videos. She wanted to understand what people feel when they watch AI content without knowing it, and what happens when they find out something they liked or loved or hated was never human at all. The result was a revealing look at how quickly audiences can spot AI, how willingly they’ll ignore it when they’re entertained, and how platforms themselves still struggle to keep up with the technology they’re hosting.

Building Janie was surprisingly simple, and that was exactly the point. Moore used ChatGPT to design Janie’s personality, looks, and style—giving her frizzy red hair, glasses, and a wardrobe inspired by the 1960s. She found a real video of a University of Alabama dorm room online and fed it into ChatGPT, which generated a still image of Janie sitting in that room. Then she used MiniMax, an AI video model, to animate that still image into moving footage. For each video, she wrote prompts that shaped how Janie would move, gesture, speak, and even what scripts she would follow. The result was a week’s worth of content that looked and sounded like a typical RushTok diary. Moore chose RushTok deliberately. It was the perfect testing ground for her experiment. The format already attracted huge audiences and encouraged daily participation, and Janie would be just one of thousands of young women posting similar videos during the same week. She wasn’t selling anything, wasn’t promoting a brand, and wasn’t trying to hurt anyone. She was simply participating, and in doing so, she became part of a shared cultural moment—except she was made of code and algorithms rather than flesh and blood.

The internet’s reaction to Janie was split from the very first video. Moore noticed a clear generational divide: younger users, especially those under 25, often clocked Janie as AI almost immediately. They spotted the subtle tells—the slightly unnatural eye movements, the overly smooth skin, the strange digital artifacts that creep into even the best AI-generated video. But older users, who made up a significant portion of the Bama Rush audience, were charmed. They left encouraging comments, wished her luck, and sent her good thoughts, completely unaware that she wasn’t real. By day two and three, however, the truth began to spread. Reddit threads popped up dissecting Janie’s videos. TikTok users created commentary and reaction videos, pointing out inconsistencies and clues. Some people even took it upon themselves to correct anyone who still believed Janie was real in the comments. Yet even after the truth was out, something unexpected happened: people kept watching. They said they didn’t care if she was AI. They thought she seemed sweet, and they wanted her to succeed. Moore realized that people were willing to embrace AI content as long as it was entertaining enough. So she shifted her strategy and started dropping subtle hints, almost like an Easter egg hunt, and watched as digital sleuths eagerly pointed out the clues in the comments.

Perhaps the most telling part of the experiment was how the platforms themselves handled it. Moore chose not to label her videos as AI-generated when she uploaded them. Of the twenty videos she posted, TikTok only flagged eight initially. By the time the experiment ended, the platform had added an “AI-generated media” warning to sixteen of the twenty videos—but that still meant four slipped through completely. Moore believes this reveals a fundamental limitation in how platforms detect synthetic content. “The platforms themselves are going to have a hard time detecting this content at scale, especially when the models are changing and the strategies are changing, and people are always gonna find ways to make this undetectable,” she explained. At the same time, she’s fascinated by the race between AI quality and human perception. It’s like two lines on a graph with different slopes. One line represents how quickly AI is approaching true human-like quality—and it’s improving fast. The other line represents how quickly humans are getting better at spotting AI—and right now, that line is rising even faster. “The average human can still tell an AI video from a truly human video,” Moore said. But she’s not sure how much longer that will be true.

When the experiment ended, Moore posted a final video revealing the truth. She explained who she was, why she created Janie, and how the whole thing worked. She expected anger or disappointment. Instead, she was shocked by the outpouring of positive comments. “I think consumer sentiment towards AI is still quite negative,” she said. “But if people feel like they aren’t being tricked, they’re being let in on it, they are being told how to do this themselves, they’re being entertained by the content that’s created, that can completely flip and the feelings can be very, very positive.” Moore’s bigger message is one of cautious optimism. She understands why people are afraid of AI—so much of the AI content we encounter online is designed to deceive, to manipulate, or to trick us into clicking. But she believes the more honest and transparent we are about how AI is used, the more it can become a tool that amplifies human creativity rather than replacing it. Janie B. was never real, but the connection people felt with her was. And that, perhaps, is the most human thing of all.