Full Breakdown
AI-Generated “RushTok” Influencer Janie B Tests Viewer Perception
By Drooid · · How we work
Core Experiment and Platform Response
Olivia Moore, an investment partner on the consumer-investing team at Andreessen Horowitz, created the TikTok persona “Janie B” as a week-long experiment to gauge how audiences react to AI-generated content. Using a real University of Alabama dorm-room clip, Moore combined ChatGPT-generated images with the MiniMax H3 video model to animate a red-haired, bespectacled freshman posting “outfit-of-the-day” videos during the annual “RushTok” sorority recruitment period. Of the 20 videos uploaded, TikTok initially flagged 8 as AI-generated; by the time of publication, the platform displayed an AI-generated media warning on 16 of them.
Audience Detection and Sentiment
The first video, posted on the first day of Alabama Rush, amassed over 120,000 views. Moore reported that younger viewers (under 25) tended to recognize the synthetic nature quickly, while many older members of the RushTok audience were initially fooled and offered supportive comments. By the second and third days, commenters increasingly identified Janie as AI, yet a portion continued to express positive sentiment regardless of the revelation. Moore observed that viewers who were “let in on it” and found the content entertaining responded more favorably.
Data & Statistics
- 20 AI-generated videos released.
- Initial TikTok AI detection: 8 flagged; later warnings applied to 16 videos.
- First video view count: >120,000.
- Audience split: younger users detected AI promptly; older users largely did not until later days.
Official Statements & Responses
She warned that platforms will struggle to scale detection as models evolve. The Alabama Panhellenic Association declined to comment on the experiment.
Verbatim Quotes
- “I would say anyone under maybe 25 immediately clocked her as AI," Moore said.” — Olivia Moore
- “We're kind of in a graph with two competing lines of different slopes. So one of them is like, how fast is AI approaching true human-like quality? And that has gotten a lot better over the past year. The other line is how fast are humans getting better at detecting AI?” — Olivia Moore
- “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,” — Olivia Moore
- “I think consumer sentiment towards AI is still quite negative. 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,” — Olivia Moore