An artist profile can have a name, a face and a catalogue of songs without a real performer behind it. Listeners may never notice. Streaming platforms are beginning to make the distinction count when recommending music.
Since September, Spotify has been rolling out ‘AI Persona’ badges. The badges appear on profiles that identify as AI personas or that Spotify assesses as likely to depict photorealistic AI-generated people.
Initial reviews focused on profiles meeting audience thresholds; artists can appeal badges applied by Spotify. The label concerns the artist’s identity, not whether their music was AI-made. Music from these profiles are excluded from editorial and algorithmic recommendations by default, unless listeners signal deliberate interest, for example by following the artist.
When supply outruns attention
Deezer, the global digital music and podcast streaming service that offers a library of over 120 million tracks, received an average of about 90,000 fully AI-generated tracks a day in June. On peak days, AI tracks made up more than half of new uploads, yet they accounted for just 1% to 3% of streams.
Deezer identified up to 85% of streams on fully AI-generated tracks as fraudulent in 2025 and excluded them from royalty payments.
An endless supply of songs is of little use if audiences struggle to find something worth hearing. A recommendation can be worth more than another thousand additions to the catalogue.
“True artist-fan connection can only be built on a foundation of trust and authenticity,” Spotify says.
Deezer labels fully AI-generated tracks and keeps them out of editorial and algorithmic promotion. Spotify’s badge deals with synthetic artist identities. An interested listener can still find the music, but is less likely to stumble across it through a recommendation.
For artists, that can mean a profile full of songs but fewer chances to reach strangers. Publishers and brands also depend on recommendation feeds. If platforms begin treating synthetic images, voices or articles differently, disclosure could affect how far that content travels.
The weight of a label
Labels can also change how music is heard. In experiments by Sarah H. Wu and Kevin J. Holmes, listeners who believed a piece to be AI-composed imagined fewer stories and reported weaker narrative engagement, even when it was written by humans. The study tested claims about composition rather than artist identities, but it showed that attribution can alter the listening experience.
In a 2025 Deezer-Ipsos survey of 9,000 people across eight countries, 97% failed a blind test asking them to distinguish two AI-generated tracks from a human-made one. Among streaming users, 73% wanted to know when fully AI-generated music was being recommended. Across all respondents, 80% supported clear labels for such music.
Europe is moving towards greater disclosure, although the legal rules are distinct from Spotify’s policy. Transparency obligations under the EU AI Act began applying to certain systems and synthetic content on 2 August 2026.
They do not require Spotify’s badge. A limited extension until 2 December 2026 applies to Article 50(2) marking and detection duties for systems placed on the market before 2 August.
An AI Persona can still release songs on Spotify. Without the platform’s recommendations, fewer listeners may discover them by chance. A badge on a profile can shape an artist’s audience–before anyone presses play.
An artist profile can have a name, a face and a catalogue of songs without a real performer behind it. Listeners may never notice. Streaming platforms are beginning to make the distinction count when recommending music.
Since September, Spotify has been rolling out ‘AI Persona’ badges. The badges appear on profiles that identify as AI personas or that Spotify assesses as likely to depict photorealistic AI-generated people.
Initial reviews focused on profiles meeting audience thresholds; artists can appeal badges applied by Spotify. The label concerns the artist’s identity, not whether their music was AI-made. Music from these profiles are excluded from editorial and algorithmic recommendations by default, unless listeners signal deliberate interest, for example by following the artist.
