Study: Fans Can't Tell AI Songs From Real Ones.

A new global study from Deezer and Ipsos suggests that most listeners can’t tell when a song has been created by artificial intelligence, raising fresh questions about trust, transparency, and the future of music discovery. The research, conducted across eight countries and involving 9,000 participants, tested listeners with a blind comparison between AI-generated tracks and a real song. Nearly everyone missed the distinction, and many participants said the experience left them uneasy.

The survey is the first large-scale international look at how consumers feel about AI-generated music. Most respondents were active music streamers, giving researchers a detailed look at attitudes among people who interact with digital platforms every day. When listeners learned that they had misidentified AI tracks as human-made, many expressed surprise, while others said they were uncomfortable with the idea that AI music could pass undetected.

Although plenty of listeners are curious about what AI can create, a sizable share said they would prefer to avoid fully AI-generated songs on streaming platforms. Some said they would skip an AI track if it appeared in a playlist, and many expressed interest in a filtering option that would let them exclude AI-made songs entirely. Still, curiosity remains strong: a large share of respondents said they would listen to AI music at least once, simply to hear what the technology is capable of producing.

One theme emerged clearly: people want transparency. A strong majority said that AI-generated music should be clearly labeled, and many listeners want platforms to disclose when their recommendation systems include AI music. A noticeable portion of respondents weren’t sure whether their streaming service had already suggested AI-made songs, illustrating how easily this content blends into existing catalogs.

The study also highlights growing concerns about how AI will affect the broader music ecosystem. Many respondents said AI threatens the livelihoods of musicians, songwriters, and composers. Others said it feels unethical for AI companies to train their models using copyrighted material without explicit permission from the artists who created it. There is also growing anxiety about how AI might affect the quality of music overall. Some listeners fear that widespread use of generative tools could lead to more generic, less creative songs flooding streaming platforms.

At the same time, the report points to one meaningful upside: discovery. Nearly half of respondents said AI has the potential to help them find more music they enjoy, whether through recommendation tools or through new types of sound that human artists might not have created on their own. But even among those who embrace AI-powered discovery, concerns about credit, compensation, and artistic integrity remain.

Deezer says it is responding to these concerns by identifying and tagging AI-generated tracks on its platform. The company says tens of thousands of AI-made songs appear online each day—now accounting for more than a third of all new daily uploads—and argues that transparency is essential as the volume continues to grow.

Deezer CEO Alexis Lanternier said the findings show that listeners value honesty about what they’re hearing and want protections for human creators. He added that the company intends to keep developing tools to distinguish AI-made content and reduce the risk of platforms being flooded with synthetic tracks.