YouTube and Deezer Move to Flag Machine-Made Songs Flooding Streaming Services

Two of the world’s music streaming services took new steps this week to tell listeners when a song was made by software rather than by people, as tracks created with generative artificial intelligence pour onto their platforms faster than anyone can listen to them.

YouTube Music said Wednesday it is adding labels to songs created primarily with generative AI, such as tracks where a model produced both the composition and the vocals. The label appears on the player page, and tapping it brings up details about the role the technology played. The labels will also appear in the main YouTube app. The company said it is working with music partners and using its own detection tools to identify such tracks, and that listeners had asked for more transparency.

YouTube is emphasizing disclosure rather than bans. Partners can designate songs as fully generated, partly generated with the artist’s own live elements, or not generated at all, and ordinary studio tools such as pitch correction do not trigger a label. Using AI does not by itself make a track ineligible to earn money on the platform.

A day earlier, the French streaming service Deezer announced a toggle that will let users exclude fully AI-generated tracks, albums and related artists from search results. The feature is expected to roll out worldwide before the end of the year and can be switched off at any time. Deezer’s chief executive, Alexis Lanternier, said demand for such a filter was clear and the company wanted to give listeners control.

Deezer has tracked the surge more closely than most. It says it now receives nearly 90,000 fully AI-generated tracks a day, more than half of all deliveries on peak days, up from about 10 percent of daily uploads when it began using its detection tool in 2025. Yet such music accounts for only about 1 to 3 percent of streams on the service, and Deezer found that up to 85 percent of streams of fully AI-generated tracks in 2025 were fraudulent, generated to siphon royalties. It excludes manipulated streams from payouts and keeps detected AI music out of its recommendations and editorial playlists.

Listeners appear to want help telling the difference. In a survey of 9,000 people in eight countries conducted for Deezer by Ipsos last year, 97 percent could not distinguish fully AI-generated songs from human-made ones in a blind test, 80 percent said such music should be clearly labeled and 45 percent said they wanted a way to filter it out.

Other services have taken a range of approaches. Bandcamp banned AI-generated tracks in January, Tidal stopped paying users who upload AI music in June, Spotify labels AI tracks without banning them, and Apple Music has said it plans to add labels by the end of the year.

The case for a lighter touch is that generative tools can also be legitimate instruments, used by working musicians to sketch ideas or produce sounds, and that disclosure respects both artists and audiences without deciding for them. Deezer’s figures suggest the more pressing problem is not taste but fraud.