Streaming platforms are now absorbing more than 100,000 AI-generated music uploads per day, according to industry reporting circulating this month. The number is a rough estimate — nobody has reliable per-day per-platform numbers — but the order of magnitude is the point. The volume is too large for human curation to filter, the per-track economics are too small for the platforms to investigate manually, and the line between “AI-assisted” and “fully AI-generated” is not legally settled anywhere.
Spotify, Deezer, Apple Music, and YouTube Music are all sitting on the same problem, and the problem is going to force them to pick a side they have been avoiding picking.
Three possible positions, increasing in spine:
- Treat AI tracks the same as human tracks. Pay the same per-stream rate. Make no labeling distinction. Let the listener decide whether they care.
- Label AI tracks. Keep them on the platform but tag them, the way some platforms tag explicit content. Let listeners opt in or opt out.
- De-prioritize or remove fully AI-generated music. Take the position that the platform exists to discover human artists, not to maximize hours-of-content-played.
The streaming platforms have so far been doing some version of option 1 by default — because pivoting to options 2 or 3 requires being able to reliably detect AI-generated music at upload time, and the detection problem is unsolved at scale. The labeling problem is harder than the labeling problem for AI-generated images because audio’s compression layer destroys the kind of metadata signal that image-watermarking schemes rely on.
The pressure to move is going to come from the artists, not the platforms. Established musicians whose per-stream payouts are getting diluted by a flood of synthetic tracks are going to start pushing — through their labels, through their managers, through their lawyers — for clearer platform policies. The labels, who collect on the artist’s behalf, are going to be the loudest voices in the room.
The interesting question for the streaming category is what happens to user behavior when an AI-generated track is also genuinely good. The most popular AI-generated tracks already cross over to human-curated playlists. Listeners don’t know. The platforms increasingly don’t either. The legal question of whether an AI-generated song can have a human songwriter for royalty purposes is in active litigation in several jurisdictions, and the answers will not be uniform.
Spotify is the platform best positioned to take a public stance. It has the largest share, the most leverage with the labels, and the most to lose if listeners start associating the platform with synthetic-track saturation. Sometime in the next six months, Daniel Ek is going to have to say something. He’d rather not.
What he says will set the policy floor for the rest of the streaming category.