Key findings: A blind survey of 9,000 listeners across eight countries found 97% failed to identify the single human-recorded track when it was presented alongside two AI-generated songs. Respondents said they were uneasy about being unable to tell the difference and supported clear labelling of AI music on major streaming platforms. Deezer reports over 50,000 AI tracks are uploaded daily — about a third of new music — and estimates up to 70% may be fraudulent; its detection tool removes confirmed AI-only tracks from recommendations. An AI-created country song, 'Walk My Walk' by Breaking Rust, also topped a US Billboard chart and has exceeded 3 million Spotify streams.
Most People Can't Tell AI Music from Human Tracks, Survey Finds — 97% Failed a Blind Test
Key findings: A blind survey of 9,000 listeners across eight countries found 97% failed to identify the single human-recorded track when it was presented alongside two AI-generated songs. Respondents said they were uneasy about being unable to tell the difference and supported clear labelling of AI music on major streaming platforms. Deezer reports over 50,000 AI tracks are uploaded daily — about a third of new music — and estimates up to 70% may be fraudulent; its detection tool removes confirmed AI-only tracks from recommendations. An AI-created country song, 'Walk My Walk' by Breaking Rust, also topped a US Billboard chart and has exceeded 3 million Spotify streams.
07:40 PM, 11/12/2025Technology
Study: Listeners struggle to distinguish AI-generated music from human recordings
A new survey conducted by Ipsos in partnership with streaming service Deezer found that the vast majority of listeners cannot reliably tell fully AI-generated songs apart from music made by humans.
Researchers tested 9,000 people across eight countries in a blind listening trial. Each participant heard three tracks — two created entirely by artificial intelligence and one human-recorded song — and 97% of respondents failed to correctly identify the human-made track.
Many participants reported feeling uncomfortable about their inability to distinguish AI-produced music from human performances. The poll also found strong public support for clear labelling of AI-generated tracks on platforms such as Apple Music, Deezer and Spotify.
'The survey results clearly show that people care about music and want to know if they are listening to AI or human-made tracks or not,' said Deezer chief executive Alexis Lanternier.
'There is also no doubt that there are concerns about how AI-generated music will affect the livelihood of artists and music creation, and that AI companies shouldn't be allowed to train their models on copyrighted material,' he added.
Deezer reported that more than 50,000 AI-generated tracks are uploaded to its service every day, representing roughly one third of all newly added music. The company estimates that up to 70% of those AI tracks appear to be fraudulent, often uploaded primarily to collect streaming revenue.
To limit the impact of such uploads, Deezer says it uses an AI detection tool to identify music created solely with artificial intelligence. Identified tracks are automatically removed from algorithmic recommendations — a step the company describes as an initial measure to prevent significant dilution of the royalty pool. Deezer also said it is considering further actions, such as updating supplier policies or removing and demonetizing offending content, but intends to proceed cautiously.
Separately, an AI-generated country song, 'Walk My Walk' by Breaking Rust, recently became the first AI-made track to top a Billboard chart in the US and has been streamed more than 3 million times on Spotify, highlighting how quickly AI-created music can gain traction on major platforms.
What this means
The findings raise questions about transparency, copyright and fair compensation for artists as AI music becomes more prevalent on streaming services. Many listeners want platforms to label AI-generated content clearly so they can make informed choices about what they listen to and support.