A large share of the music now uploaded to streaming services is AI-generated. In April 2026 Deezer said it was receiving about 75,000 fully AI-generated tracks a day – roughly 44% of everything uploaded to it, up from around 10,000 a day a year before. Those tracks made up only 1 to 3% of what people actually streamed. Most of them get few or no plays. The damage comes from the minority that do: AI songs have charted, and every stream and chart place they take is exposure and income lost to human writers.
AI generated music hits functional and background music hardest – workout playlists, ambient tracks, the music under adverts and on television – where listeners may barely notice who made what. That is the easiest place to swap human recordings for cheaper AI ones, and some platforms are already moving that way. YouTube, for one, now offers a tool that replaces music flagged for rights problems with AI-generated music instead.
Whether a track is labelled as AI matters more than it might sound. If you cannot tell what is human-made, you cannot choose to back human music, and platforms have not been consistent about labelling. Some services detect AI tracks – usually by spotting tell-tale artefacts left in the audio – and tag them, though the detection is imperfect and tools already exist to scrub those artefacts out. Others have done little.
For writers and composers this is more an advocacy issue than a legal one. What they can realistically demand is clear, consistent labelling, so listeners know whether a track is human-made and can choose accordingly. Listeners want this too: in a Deezer-commissioned survey, 80% said fully AI-generated music should be clearly labelled. Deezer already tags the AI tracks it detects, and is so far the only major platform doing so.
