Could you tell whether a song was created by a human musician or generated by artificial intelligence simply by listening to it?
For most listeners, apparently not.
In a blind listening test commissioned by streaming platform Deezer and conducted by Ipsos, 97% of participants failed to correctly distinguish fully AI-generated music from human-created music.
The study involved 9,000 people across eight countries and comes as synthetic music is arriving on streaming platforms at an extraordinary rate.
AI music is becoming a major part of new uploads
In July 2026, Deezer reported that fully AI-generated tracks had exceeded 50% of all new music delivered to its platform on peak days in June.
The company reported an average of around 90,000 fully AI-generated tracks being delivered to the platform every day during the month.
That does not mean half of what people are actually listening to is AI-generated. The volume refers to newly delivered music rather than overall listening.
Nevertheless, the scale of new synthetic material creates a significant challenge for streaming services, rights holders and listeners.
Humans struggle to hear the difference — but software can help
There is an important distinction between human listening and technical detection.
The Deezer-Ipsos study suggests that listeners frequently cannot determine whether music is AI-generated by sound alone. Technical detection systems, however, can analyse characteristics within audio files that may not be perceptible to human listeners.
Deezer says its AI music detection technology can identify fully AI-generated music with 99.8% accuracy.
This means the issue is not simply that the music industry has no way of recognising AI music. The larger challenge is identifying, managing and labelling synthetic content reliably at enormous scale as generative tools continue to improve.
Listeners increasingly want transparency
The same Deezer-Ipsos survey found strong demand for clearer disclosure. Eighty per cent of respondents said fully AI-generated music should be clearly labelled for listeners.
That may become one of the most important questions surrounding AI music: not whether synthetic music should exist, but whether audiences should always know what they are listening to.
The transparency issue now extends beyond streaming platforms. Musical instrument and accessory companies are also confronting questions about generative content after D’Addario acknowledged that generative AI had been used in music for a controversial product demonstration.
Streaming services must also consider recommendations, royalty distribution and fraudulent streaming. Deezer has taken steps to identify and restrict fully AI-generated material within parts of its platform.
AI music is also creating major copyright disputes
As synthetic music becomes more convincing, another question is growing alongside it: what material was used to train the systems capable of producing it?
That issue is now moving through the courts. Sony Music Publishing and Warner Chappell have sued AI company Anthropic over allegations involving copyrighted songs and sheet music used in AI training.
The case highlights how the AI music debate now covers several separate issues at once — whether listeners can recognise synthetic music, whether it should be labelled, how royalties should be handled and whether copyrighted material can legally be used to train generative systems.
The line between human and AI music is getting harder to see
Generative AI is increasingly capable of producing vocals, arrangements and complete recordings that can sound convincing on first listen.
But music creation is also unlikely to remain a simple choice between “human” and “AI”. Musicians may use AI for individual parts of a production while continuing to compose, perform and record other elements themselves.
For listeners, platforms and musicians, transparency may therefore become more useful than attempting to judge a recording purely by ear.
