On Wednesday 14 October, YouTube Music began gradually showing a label directly on the player page for tracks created primarily with generative AI. How many people hear such a track, though, depends mainly on recommendations, and Deezer automatically removes the tracks it detects as AI from those. A day earlier, Deezer announced that by the end of 2026 it will add a toggle that lets listeners hide fully AI-generated music from search results as well.
On peak days, fully AI-generated tracks make up the majority of new music on Deezer, but only a fraction of what people listen to. Since July 2026, Deezer has said it receives nearly 90,000 of them a day and that they account for barely 1 to 3% of streams. It puts this down to removing every track detected as AI from recommendations and editorial playlists, which drastically limits their organic reach.
Where platforms and the AI Act draw the line for labelling AI music
Deezer's detection currently focuses on content created entirely on generative AI platforms, and the toggle also targets fully AI-generated music. Tracks where AI is only a limited aid within a broader human creative process are treated differently by Deezer. On YouTube, one example of a partly AI track is a song where the artist has AI generate one or more layers, such as a bassline or strings, and then records their own live vocals and instruments over them. YouTube puts standard pitch correction such as Auto-Tune, and AI mastering, in the no-AI category. According to YouTube, an AI-generated music video for a song recorded entirely by humans makes it a partly AI track.
On YouTube, a track's category is set first and foremost by whoever delivers it. Music partners can declare in the metadata whether a track is fully AI, partly AI or AI-free, using the categories Fully Gen AI, Partly Gen AI and No Gen AI. They submit the declaration via DDEX, the industry metadata standard, or via a spreadsheet template. Those who upload on their own through YouTube Studio can flag AI use in the attributes section. YouTube leaves the interpretation of these definitions to partners. When a partner says nothing about AI, YouTube relies solely on its own signals and may itself designate the track as fully or partly AI.
The obligation to label content comes from the transparency rules of the EU's Artificial Intelligence Act, the AI Act. The rules apply from 2 August 2026. Providers of AI systems that generate audio, images, video or text must mark their outputs in a machine-readable format. A visible disclosure for audio and video is required only from whoever uses AI to produce a deepfake.
By a deepfake, the law means content that resembles existing persons, objects, places, entities or events and could falsely appear to a person to be authentic. For an evidently artistic work, an appropriate disclosure is enough. It must not hamper the enjoyment of the work. The “Fully AI-Generated” icon that the European Commission offers for labelling lists fully AI-composed music among its examples. Yet not all AI content has to be labelled.
Recommendations and payouts for AI tracks
Deezer introduced AI music labels back in June 2025, and since then over 14 million albums have been tagged. Between January 2025 and August 2026, it removed over 42 million AI tracks from algorithmic and editorial recommendations.
According to Deezer, up to 85% of streams of fully AI-generated tracks in 2025 were fraudulent. The company excludes manipulated streams from payouts.
On payouts at YouTube, TechCrunch writes that “AI involvement alone doesn’t make a track ineligible for monetization”. For videos, YouTube stated back in May that a label alone does not change recommendations or eligibility to earn money. Whether the same applies to tracks is not spelled out in either YouTube's blog post or its help page.
Deezer links recommendations to payouts explicitly. Removing AI tracks from recommendations and editorial playlists is, it says, a first step towards making sure they do not significantly dilute the royalty pool. The company says further steps, including changes to its supplier policy, depend on careful consideration. Back in July, however, Deezer announced that it would systematically take down AI tracks used for streaming fraud. A track it detects as AI does not get into recommendations, and if no listener plays it for at least six months either, Deezer takes it down.
Comments