Deezer pulls payouts on most fully AI tracks and will sell its detection tech

Deezer has moved from labeling to financially penalizing fully synthetic music after internal analysis found massive volumes of AI-generated uploads and widespread manipulation. The company says it identified more than 13.4 million AI-created tracks on its service during 2025, with contributions running at roughly 60,000 new AI songs every day — a torrent that now represents about 39% of daily content deliveries. Despite that supply surge, listening to AI-only material remains comparatively low, accounting for roughly one to three percent of total streams; Deezer’s internal signals indicate most plays tied to those tracks are artificially inflated. To blunt the problem, the platform is flagging purely synthetic recordings, removing them from recommendation systems and playlists, and withholding royalties from streams it deems fraudulent — a measure the company reports has affected as much as 85% of plays linked to fully AI-generated tracks in some months. For context, Deezer estimates that fraudulent streaming across its entire catalog averages around eight percent, making the concentration in synthetic music a distinct threat to fair payouts. The firm has begun packaging the detection technology it developed and tested with rights organizations for licensing to other services, positioning the tool as a potential industry standard for transparency and anti-fraud enforcement. That commercial strategy is coupled with intellectual property steps: the company has filed patents on signature-based methods designed to separate synthetic audio from human performances. The implications cut two ways: platforms that adopt detection could better preserve compensation for human creators, while the market for low-cost synthetic releases may lose its profitability if demonetization becomes common. Rights groups and industry studies add urgency; one analysis cited suggests creator income could suffer materially from unregulated synthetic content, projecting substantial revenue exposure through the end of the decade. Regulatory pressure and cross-industry coordination are likely to follow if major services adopt similar stances, and metrics such as recommendation exposure, playlist inclusion, and royalty allocation will become battlegrounds for platform policy. For labels, publishers and independent artists, broader deployment of detection tools could restore some trust in streaming economics but will also raise new questions about false positives, appeals, and the standards used to certify content provenance. For the developers of generative audio, the emergence of monetization controls and detection patents will force a reconsideration of business models and distribution channels. As the industry tests interoperable defenses against exploitation, the balance between innovation and the economic viability of creators will shape the governance norms that eventually emerge. Deezer’s actions turn theoretical worries about synthetic music into operational policy, inviting other stakeholders to choose between cooperation, competition, or litigation in the months ahead.