After a thousand live-reviewed AI music submissions, certain patterns are no longer guesses. They're data. Here's what the queue has actually taught us about what lands with live audiences — and what doesn't.
The single biggest predictor
It isn't genre and it isn't production quality. It's whether the first thirty seconds give the listener a reason to keep paying attention. Tracks that earn the room early earn it for the whole song. Tracks that meander for two minutes lose the room and don't get it back, no matter how strong the chorus is.
Vocals are the make-or-break
Vocal-led tracks outperform instrumentals roughly two to one in tip volume, but only when the vocal feels like it means something. Generic AI vocals — pleasant but emotionally flat — actually underperform instrumentals. The implication is clear: if your vocal isn't doing real emotional work, the track is better off without one.
Genre breakdown
Across the first thousand submissions, the strongest-performing genres on live shows have been alt-pop, afrobeats, hyperpop, slow electronic, and country. The weakest have been generic EDM, lo-fi without a point of view, and "epic" cinematic music with no song underneath it.
Hybrid workflows win
Tracks where the artist used AI for the instrumental and sang the vocal themselves had the single highest promotion-to-AI-Radio rate of any category. We expect that to be the dominant workflow on the platform by year-end.
What it means for new creators
If you're submitting for the first time: open with a real moment, treat the vocal as the most important part of the track, and pick a genre that has a point of view. Submit your work here.