Tilbury Built Three AI Influencers With $99. Two Flopped. The Workbook Carried It, Obviously.
Mark Tilbury spent $99 and seven days cloning the Lil Miquela and Mother Satori playbook using Higgsfield plus Claude to generate three AI influencers named Amos, Vivian, and Sienna. The winning character sold a $12 workbook that produced $157 profit across 23 sales. No face, no voice, no personal brand required.
This demonstrates the audience-asset inversion principle. You are not building a personality, you are building an audience that can be redirected to any product at any time. The AI influencer is merely a customer acquisition channel. The actual asset is the distribution list and the digital product at the end of it.
Mark Tilbury ran the experiment on YouTube with a $99 budget constraint, proving the concept with 23 sales generating $157 profit. Three characters tested against each other, one survived.
- Open Claude and describe a fictional character persona with a specific niche like health or fitness. Ask it to write a week of daily captions and a video concept for each day. Expected outcome: a content calendar with seven posts and visual descriptions.
- Use a free trial of an AI image or video generation tool like Higgsfield to create a consistent character across five to seven images or short clips using the same character description each time. Expected outcome: a set of visuals that look like one person posted them.
- Create a simple one-page digital product in Google Docs, export it as a PDF, and list it on Gumroad for $5 to $12. Expected outcome: a live product link you can put in any bio.