Qobuz labels AI music after finding 60% of streams on AI tracks fraudulent
The French streaming service has removed millions of tracks from search and barred identified synthetic releases from recommendations and editorial promotion.
Qobuz has begun displaying an in-app label on releases containing music it has identified as AI-generated, pairing the transparency measure with a crackdown on suspected streaming manipulation. The French streaming and download service said 60% of streams on those AI-identified tracks were deemed fraudulent and would be excluded from royalty payments to rightsholders.
The tag, announced Sept. 24, appears at album or release level rather than on individual tracks. Qobuz said the purpose was to let listeners identify AI-generated material clearly and make informed choices.
AI-identified tracks accounted for 0.38% of all streams on Qobuz. The company attributed that relatively small share to its human-led discovery model, which keeps such content out of its discovery journeys.
Qobuz has removed more than a third of AI-identified albums with no listening activity from its search engine, covering 4 million tracks. The releases remain available when a listener searches their titles directly, under the platform’s updated AI Charter.
The service does not promote identified AI-generated material in editorial sections including Our Selections, New Releases, Qobuzissimes, Albums of the Week and playlists. It also excludes the content from algorithmic recommendations.
The measures build on an in-house identification system first announced in late February, when Qobuz said visible labels would follow in the coming months. Its September AI Charter says the platform supports AI as a creative tool for artists, provided a human validates the production and retains final responsibility. But it opposes industrially produced AI content with no human creative intervention when used to flood catalogues or defraud royalty systems.
Qobuz acknowledged a technical limit: no reliable method can identify all AI content with certainty, particularly where humans have partly intervened. Artists who believe a release has been wrongly labelled can contact their distributors, which handle those requests, while listeners and others can submit suspected unlabeled AI content through a reporting form.
The company is also integrating a DDEX metadata field intended to identify AI-generated tracks automatically when rightsholders declare them. Deployment across labels, distributors and aggregators is under way.
Qobuz’s figures add to evidence of a broader fraud problem surrounding synthetic music. Deezer said in January that AI-generated tracks made up as much as 3% of its total streams in 2025 and that up to 85% of streams on those tracks were fraudulent, depending on the month. Deezer removes fully AI-generated music from algorithmic recommendations and editorial playlists, and excludes detected fraudulent plays from payment.
As covered in September, Italy’s music industry has similarly warned that AI-generated tracks and bots can divert money from the streaming royalty pool. Qobuz said it will continue to revise its AI Charter as the technology and regulatory environment change.