Deezer says 43 percent of users switching to it from a rival service arrive with AI generated tracks already in their playlists, and the new tool lets any listener reproduce that finding against their own account.
You can now run a free AI music detector against your Spotify or Apple Music playlist without leaving the service. Deezer, the French streaming company that has spent two years building its own AI music classifier, announced on Thursday that the tool is now free for users of 20 major streaming platforms, including Spotify, Apple Music, YouTube Music, and Amazon Music. The scan runs in the background once a listener connects an account, and it flags tracks that look likely to be AI-generated. It does not remove them.
That last part matters. The scanner is a transparency layer, not a filter. Tracks get tagged as probable AI, and the listener decides what to do next, whether that is skipping them, removing them, or just knowing the catalog better. Deezer positions the move as giving listeners a clearer view of what is being recommended to them on rival services, and CEO Alexis Lanternier has framed it as a question of honesty in music discovery.
The scale of what the scanner is built to expose is what makes the announcement newsworthy. Deezer, which acts as both a competitor to Spotify and Apple Music and a would-be neutral arbiter in this fight, has been publishing its detection numbers for more than a year, and the recent figures are stark: roughly 75,000 fully AI-generated tracks submitted to streaming platforms every day, more than 44 percent of daily upload volume on Deezer's platform classified as fully synthetic, and more than 13 million AI tracks tagged on Deezer alone in 2025. Those figures are vendor-reported, drawn from Deezer's own pipeline, and they have not been independently audited.
The number that lands hardest is about listeners, not uploads. Deezer says 43 percent of people who switch to Deezer from another platform arrive with AI-generated tracks already sitting in their playlists, tracks the user did not pick and, in most cases, did not know were synthetic. That is a finding about the recommendation and editorial machinery of mainstream streaming, not about any one listener's taste. The scanner, in effect, lets anyone reproduce that finding on their own account.
You can see the same saturation in the charts. A song called "Walk My Walk" by the artist Breaking Rust, an AI-generated country track, reached the top of Billboard's Country Digital Song Sales chart. IngaRose's "Celebrate Me," another AI-made record, hit number one on the U.S. iTunes chart. These are not curiosities. They are evidence that synthetic tracks are reaching the discovery surfaces listeners trust, the editorial and algorithmic places where most people still assume a human is curating what they hear.
The scanner itself is straightforward to use. A listener connects a Spotify, Apple Music, YouTube Music, or Amazon Music account, and Deezer reads the playlist data, runs each track through its classifier, and returns a list of likely AI songs. The result is a percentage and a label, not a removal. The model is probabilistic, and the company has not published precision or recall numbers for the public tool, which is one of the most important caveats for any reader planning to act on the results.
Two things are worth keeping in mind before drawing conclusions. First, Deezer is a competitor to the very platforms it is now helping listeners audit, and it has been the most aggressive major streamer on AI tagging. Lanternier's framing of this as a transparency play is also a positioning play, and any reader should weight it accordingly. Second, the most important outside voices in this story, independent music technology researchers, performing-rights organizations, and the human artists whose work sits next to synthetic copies in the same playlists, have not weighed in on this specific rollout in the available reporting. The data is real, but the interpretation is still mostly Deezer's.
What the tool does give a listener is a way to test the claim against their own listening history. Run it against the playlist you have been adding to for a year, the one you share with friends, the one that surfaced from a public radio editorial pick. The interesting question is not whether the scanner finds something. It almost certainly will. The question is what streaming services owe listeners once it does.