The top 3% of YouTubers take 90% of the money. AI is about to make it stop working.

Numbers first so this isn't vibes. Top 1% of creators now take 21% of all creator payments, up from 15% three years ago. Top 10% take 62%. On YouTube the top 3% of channels take about 90% of net creator earnings. Half of working creators clear under $10k a year. The standard explanation is superstar economics: when it costs nothing to copy a performance, everyone watches the best one, so the best one gets everything. That's been the textbook answer since 1981 and it's why nobody in the industry treats those numbers as a problem. I think the textbook answer has quietly stopped being true, and AI is the reason. Superstar economics needs two things: the top performer has to be meaningfully better, and the audience has to be able to find out. The second one was always the platform's job. The recommender decides what "comparable" means, and for a decade it has answered that question the lazy way: comparable to what already has watch time. A new channel with comparable content and zero history is invisible not because it lost a contest but because the contest was never run. Platforms could tolerate that when content was scarce enough that the top 3% were obviously better. What changes with AI is the supply side. When a solo creator with a generation stack can produce something 80% as good as a studio, at 1% of the cost, the pool of "comparable" content explodes, and the recommender's bias toward incumbents stops looking like taste and starts looking like a mispriced asset. A market that keeps paying 90% of the money to 3% of suppliers, while comparable supply piles up unpriced, is a market that's about to get arbitraged. The arbitrage is already visible at the edges. Platforms are now paying for provenance (Reddit sells verified human posts to AI labs for $60–70M a year per buyer) because verified human-made is the scarce input now, not content. YouTube's own AI policy carves out exactly one protected category: a human on camera with original commentary. They're telling you what they think the scarce asset is. It isn't the top 3%'s polish. It's the person. So the prediction: the first platform that builds a recommender that actually runs the contest, that tests new supply against incumbents on content rather than on history, takes the long tail from everyone else, because that's where the underpriced inventory is. And if none of the incumbents build it, the capital flows around them, the same way it flowed around cable. Where I could be wrong: maybe audiences don't want comparable, they want famous, and fame is the product. In that case the 3% keep the money and AI just makes the 97% cheaper to ignore. I'd like to hear the case for that, because I don't think it survives the supply shock, but I've been wrong about markets before.