Music streaming services like Deezer, Spotify, and YouTube Music have become the primary way people discover and listen to music. While this convenience benefits listeners and artists alike, it also opens the door for AI-generated content to infiltrate playlists and charts. Deezer, a French audio streaming platform with over 16 million active users, has been quietly developing technology to identify and manage AI-created music. In a recent update, Deezer disclosed that nearly half of all new music uploads to its service are now generated by artificial intelligence, and the majority of those streams are fraudulent—often generated by bots or other automated systems.
According to Deezer, the proportion of AI-generated uploads has soared to 44 percent year-to-date, which translates to approximately 75,000 new AI tracks every single day. This staggering volume raises questions about the future of music streaming economics, copyright, and the very definition of creativity. Deezer's CEO Alexis Lanternier noted, 'Thanks to our technology and the proactive measures we put in place more than a year ago, we have shown that it's possible to reduce AI-related fraud and payment dilution in streaming to a minimum.'
The company says it has developed proprietary detection algorithms that can identify AI-generated audio with a false positive rate of less than 0.01 percent. This technology is already being licensed to third parties, indicating that Deezer sees a growing market for AI detection in the music industry. Unlike some competitors, Deezer explicitly labels AI-generated content, and it does not include such tracks in editorial playlists or algorithmic recommendations. As a result, AI music accounts for only about 1–3 percent of total streams on the platform, despite representing a huge chunk of new uploads.
Deezer's survey found that an astonishing 97 percent of listeners could not tell the difference between AI-generated songs and human-made ones when presented with a blind test—two AI tracks and one real track. This suggests that the quality of AI music has improved dramatically, making detection essential rather than relying on human ears. The implications are far-reaching: if AI can create indistinguishable pop songs, the entire royalty system may need reform to ensure human artists are compensated fairly.
The fraudulent nature of most AI music streams is a critical point. Deezer only pays royalties when a real person listens to a track, so it has moved to demonetize roughly 85 percent of AI music streams. This means the vast majority of AI uploads are essentially spam, designed to game the platform's payment system. Botnets or automated scripts can stream thousands of AI-generated tracks daily, siphoning money from a limited royalty pool that would otherwise go to legitimate artists. Deezer's approach—detecting and demonetizing rather than outright banning—is pragmatic, given the sheer volume of uploads.
The growth of AI music is accelerating alongside the broader AI industry. Models like Google's Lyria 3, Suno, and Udio allow users to generate full-length, broadcast-ready songs in seconds. Google now lets Gemini users generate complete tracks, up from 30-second snippets just a few months ago. These mainstream platforms typically embed watermarks—such as Google's SynthID—to flag AI-generated content. However, Deezer says it is becoming increasingly easy to strip those watermarks from audio files or generate music using custom models that never include them in the first place. As AI inference costs drop, the creation of 'musical AI slop' becomes cheaper and more accessible.
Industry observers note that this trend is not unique to Deezer. Spotify and other major streamers have also seen a surge in AI uploads, but they have been less transparent about their detection methods. Some artists and labels have expressed concern that AI music could dilute the value of human creativity and reduce the discoverability of emerging talent. Meanwhile, streaming services face a dilemma: they want to offer vast catalogs, but unchecked AI uploads could overwhelm human curation and inflate costs.
Deezer's proactive stance positions it as a leader in the fight against AI-generated fraud, but the battle is far from over. As generative audio models continue to improve and proliferate, the rate of AI uploads is likely to climb beyond the current 44 percent. The company's detection technology will need constant updates to keep pace. For now, Deezer is betting that identifying and flagging AI music—rather than blocking it entirely—is the sustainable path forward, allowing human listeners to choose while preventing abuse of the royalty system.
The broader context is that AI-generated music is a double-edged sword. On one hand, it democratizes music creation, allowing anyone to produce a song without musical training. On the other hand, it threatens the livelihoods of professional musicians and introduces new forms of fraud. Deezer's data suggests that the vast majority of AI uploads are not creative experiments but calculated attempts to exploit the streaming economy. Without robust detection and monetization controls, streaming platforms could become flooded with synthetic music that drains revenue from real artists.
Looking ahead, the music industry will need collaborative solutions. Watermarking standards, shared databases of AI-generated tracks, and regulatory frameworks may help. Deezer is already licensing its detection tech to third parties, hoping to create a unified front. But as the CEO acknowledges, the pace of AI advancement means no single company can solve the problem alone. For now, Deezer's numbers—44 percent of new uploads, 85 percent of AI streams demonetized—serve as a stark warning for the entire music streaming ecosystem.
Source: Ars Technica News