A global survey highlights growing concerns about authenticity in digital music as platforms tighten rules around misleading audio content
A newly released Deezer study, published on November 12, reveals that the public still struggles to recognize music fully generated by artificial intelligence. Conducted by Ipsos with 9,000 people across eight countries, the research showed that 97% of participants failed to correctly identify any of the three tracks created entirely by AI.
The result surprised 71% of respondents, and 52% reported discomfort upon learning they couldn’t distinguish human vocals from synthetic ones. Participants also expressed support for clear labeling and fair compensation for artists whose work is used to train AI models. So far, Deezer remains the only streaming platform capable of detecting and tagging 100% AI-generated content.
Days after the study was released, the industry saw a real-world example of this lack of clarity. On October 28, “I Run,” uploaded by the enigmatic artist HAVEN, racked up more than 13 million Spotify streams before being removed due to suspicion of artist impersonation. The vocals were compared to Jorja Smith’s, who denied any involvement.
Producer Harrison Walker later identified himself as HAVEN, claiming the vocals were his own but heavily processed using AI tools. An engineer involved in the track noted the use of Suno, a generative AI platform.
The removal reinforces the growing effort among streaming services to enforce stricter policies against content that may mislead listeners or mimic real artists without transparency.
