Deezer Study: 97% of Listeners Can't Tell AI Music From Human-Made Tracks

French streaming service Deezer has revealed that AI-generated tracks now represent 44% of all new daily uploads on its platform. Roughly 75,000 such compositions land on the service every day — exceeding 2 million per month. A parallel study of 9,000 participants across eight countries found that 97% of listeners could not distinguish AI-created music from songs made by humans.

Why This Matters

The explosive growth of AI-generated content on music platforms threatens the livelihoods of real artists and erodes trust in streaming ecosystems. If virtually no listener can identify synthetic tracks, the industry faces a massive royalty fraud problem without robust technological and regulatory safeguards. Deezer's findings serve as a wake-up call for the entire music business to take collective action.

Massive Uploads, Minimal Engagement

Despite the staggering volume of AI uploads, actual listener interest remains negligible — these tracks account for just 1–3% of total streams. The gap between uploads and plays is largely explained by fraud. Deezer's proprietary detection system flagged 85% of AI track plays as artificially inflated. The platform subsequently excluded those streams from royalty calculations.

Deezer launched its AI Detection tool in January 2025, becoming the first streaming service to label neural network-generated content. The tool claims 99.8% accuracy and identified over 13.4 million AI tracks in the platform's catalog over the past year.

Deezer CEO Alexis de Lanternnier stated that AI music is no longer a fringe phenomenon and called on the entire industry to unite in protecting artist rights and ensuring transparency for fans. Starting April 20, the platform also stopped offering high-quality versions of AI-generated compositions.

Listeners Can't Hear the Difference

Deezer's research among 9,000 respondents from eight countries produced striking results: 97% of participants failed to identify which tracks were created by humans and which by algorithms. Yet users themselves are aware of the issue:

• 80% of respondents support mandatory labeling of AI-generated content;
• 73% want to know when a service recommends fully AI-generated music;
• 52% believe AI tracks should not appear in main charts alongside human-made songs.

AI Bots Stealing Artist Identities

The threats from AI music extend beyond stream manipulation. The Guardian previously reported the story of jazz pianist Jason Moran, who discovered an album on Spotify released under his name that he had never created. Moran does not use the service and has never uploaded music to it. Upon investigation, he found a profile bearing his name with several albums from a label he had previously worked with, plus an unfamiliar indie-pop EP.

Moran noted that the recording contained no pianist whatsoever and bore no resemblance to his work. The investigation revealed that AI bots were impersonating real musicians behind such profiles.

Spotify acknowledged the problem. In September of last year, the company reported removing over 75 million "spam tracks" within a 12-month period and announced stricter policies against artist impersonation.

Music Generation Tools Keep Expanding

The surge in AI content is fueled by increasingly accessible generation tools. In February 2026, Google released Lyria 3, a model capable of producing audio from text prompts, photos, or video. The growing availability of such technology continues to lower the barrier for mass-producing synthetic music, intensifying pressure on streaming platforms and rights holders alike.