A Venture Capitalist Spent $100 Creating A Fake Sorority Girl. TikTok Didn’t Notice.
A Silicon Valley venture capitalist, Olivia Moore, conducted an experiment creating "Janie," an AI-generated virtual influencer. Using AI tools, Moore spent 30 minutes daily for a week, generating videos of Janie's fictional sorority rush experience. Janie quickly amassed thousands of views and followers on TikTok, demonstrating the ease and low cost of creating engaging synthetic content. Notably, TikTok's AI labeling was slow and didn't impact performance, suggesting algorithms prioritize engagement over authenticity. This experiment highlights the collapsing barriers to virtual personas, questioning the economic value of human influencers and the future of the creator economy. It also raises concerns about platform governance and the potential for undisclosed AI-generated content.
Digital influence has become the currency of our times. While artificial intelligence is often used to that end, a recent experiment by a Silicon Valley venture capitalist is giving voice to an important debate: If a fictional character can generate thousands of views for the price of a textbook, what is the actual economic value of a human influencer?
The case of “Janie”, an AI-generated student at the University of Alabama, is arguably one inflection point in the bigger consumer technology story, illustrating that the technical barriers to creating a credible virtual persona may have just come closer to collapsing entirely.
Olivia Moore, a partner on the investing team at the venture capital firm Andreessen Horowitz (a16z), recently published a post on her social media experiment. Moore engineered “Janie,” a red-haired 19-year-old, using a single image generated by ChatGPT. Over the course of a week, Moore spent approximately 30 minutes per day using tools like MiniMax and Grok Imagine to animate Janie’s daily life, specifically targeting the “Sorority Rush” niche, and the results were statistically anomalous for a new account. Janie’s videos — depicting rush outfits and dorm room tours — amassed thousands of views within seven days.
“Like many of the 2,500 prospective new members (PNMs), she posted outfit-of-the-day videos (OOTDs), breathless recaps of her house visits, and clips fumbling trending dances,” Moore stated in her post. “Janie created a TikTok account specifically for college - she’d never posted before - but within a week, she had 1,300 followers and was averaging tens of thousands of views per video.”
“I don’t think we’re doomed – but we are about to be outnumbered. The cost curve on frontier intelligence is only going down. ‘Supply’ of this kind of content is about to explode, and Janie is just an early entry in a category that’s evolving faster than the discourse around it,” Moore noted in her analysis of the experiment’s feedback.
One of the most notable data points to emerge from the Janie experiment pertains to the thorny issue of platform governance. Despite Janie being entirely AI-generated, the content was not flagged by TikTok’s moderation systems immediately. Moore reported that out of 20 videos posted, TikTok’s automated systems eventually applied an “AI-generated” label to eight out of those 20 videos.
“It had no visible effect on performance, or on the comments,” Moore stated.
While platforms are under pressure to identify synthetic media, the consumer appetite for content is undiminished: the algorithm did not punish Janie for being fake; instead, it rewarded her for being engaging. The user interface of social media encourages the suspension of disbelief, which is exactly what makes it a fertile ground for AI-generated personas.
The Janie experiment concluded with a reveal that, per Moore, it was met with a largely positive response from the audience.
“The overwhelming majority of feedback was positive – commenters compared it to watching Black Mirror, asked me to ‘do another one’ next year, and requested I create a TikTok to share more educational tech content,” Moore stated. The audience may not have felt “tricked” as much as they felt entertained by the meta-narrative of the experiment.
“Today, Janie’s viewers could answer ‘is she real?’ by zooming in on her hands or her dresser. Soon, they may not be able to (conclusively) answer it at all. Which makes the more interesting question: does it matter?”
The Janie experiment is also a stress test for Section 230 and platform policies. If a single individual can create a viral persona for $100, the economic model of the more than $250 billion creator economy comes a lot more into focus. Also, the consumer technology risk is in the asymmetry of information. While Moore eventually disclosed the synthetic nature of Janie, the tech also exists for actors who will not.
“Janie did not generate herself. Every rumpled outfit, mistimed comment and bad decision came from a human. That makes me think disclosure may ultimately be good for AI, not just for audiences. When creators hide the tools, people experience AI as a trick. Showing the work creates the possibility of a different relationship: A person had an idea, made choices and used AI to bring it to life.”
The Janie experiment, then, is as good as any case study on post-authentic media. In the attention economy, “synthetic” is a feature that audiences are increasingly proving open to engage with, provided the narrative quality is high, and usually, disclosure and ethical standards are met.
“I’d happily buy a reproduction of a painting – as long as I wasn’t expecting the original,” Moore stated.