The web scanner accepts exports from roughly 20 streaming services and works without a Deezer account, but it ingests the playlists you upload, and the company has not published its false positive rate for human artists.
Deezer, the French streaming service that has spent years trying to compete with Spotify and Apple Music, just made a quieter kind of argument for relevance: it has built a free, web-based scanner that checks your playlists for AI-generated tracks, and it works whether or not you have a Deezer account.
That positioning is the story. Spotify, Apple, and YouTube Music all host a growing share of synthetic music, and none of them ship a comparable listener-facing detector. Deezer, a service that has never been the default for most English-language listeners, is the one offering a tool to tell the rest of us how much of our own music is synthetic. The bet is that disclosure, even rough disclosure, becomes a wedge.
The scanner accepts playlist exports from roughly 20 services, including Spotify, Apple Music, YouTube Music, Tidal, SoundCloud, and Pandora. It returns a percentage breakdown of likely AI tracks and lets users click through to flag or remove them. Under the hood, Deezer says it is the same detection pipeline it already uses internally to label what it describes as hundreds of thousands of AI-generated tracks on its own service. That scale figure is a company claim, not an audited number, and Deezer has not published the model's architecture, training data, or false-positive rate.
The hands-on part of the story is more useful than the marketing. In a test run by CNET against a sizable Spotify library, AI-generated tracks were hard to find, which is consistent with reporting that synthetic music clusters in niche genres, generic stock-style mood music, and unsigned SoundCloud uploads rather than in the editorial playlists most listeners follow. A scanner that is honest about what it cannot find is more useful than one that inflates the threat. It also means the percentage a user sees may understate reality, depending on what they actually listen to.
The other trade-off is the one users are most likely to skip past. The site works without a Deezer login, but importing a playlist from another service pipes the track list, the service name, and timestamps into Deezer's systems. After the scan runs, the site offers to build a Deezer library from the imported playlist and to keep it in sync going forward. There is no paid tier and no premium gate on the detector itself, which is the structural point. The product is a customer-acquisition funnel with a transparency wrapper.
What the tool does not yet do is the part that matters most for independent human artists. Deezer has not disclosed what happens when a human-made track is misclassified as AI, or shared a benchmark for accuracy on unsigned or human-vocal music. For an artist whose SoundCloud upload is wrongly flagged, a tool like this is not neutral. It is a signal that listeners, playlist curators, and downstream platforms may use to deprioritize or remove their work. Surfacing uncertain results is a reasonable default. Making the false-positive rate public would be a better one.
What to watch next: whether any of the larger streamers ship their own listener-facing detectors, whether the major labels converge on a common labeling or disclosure rule for AI-assisted uploads, and whether the distributors that sit between artists and streaming services start requiring any kind of AI-use attestation at the point of upload. For now, the burden of proof sits with the listener, and Deezer's new scanner is the first mainstream product to make that explicit.