aipopdb

22 Nov 2026  ·  10,178 works · 1,193 news indexed

News / Duke study says Spotify's Release Radar surfaces unlabelled AI artists

Duke study says Spotify's Release Radar surfaces unlabelled AI artists

AIPopDB Newsroom · via techpolicy.sanford.duke.edu · · 2 min read · Comment

Read the original ↗
Image for Duke study says Spotify's Release Radar surfaces unlabelled AI artists

Image: Duke Sanford Tech Policy Program

Researchers at Duke's Sanford School have published a study and a blog post arguing that Spotify listeners are being steered towards AI-generated music without being told. The Tech Policy Program team tried to spot AI artists the way an ordinary listener would, using only the app and public information, and concludes that voluntary disclosure is not working.

“This approach to labeling is ineffective and does not promote a culture of accountability in the music industry or creative economy.”

Duke Tech Policy Program blog post by Liz Sparacino, on voluntary artist labels, in a Duke blog post · techpolicy.sanford.duke.edu ↗

Five new Premium accounts were set up, each tuned to a single genre, by a team of four undergraduates and one graduate student. Over their listening they met 728 artists, 76 of whom needed closer scoring. The checks looked for open AI disclosure, absence from Bandcamp (which bans music made largely with AI), a very high release volume, AI-style imagery, a thin online footprint and no history of live shows; anyone with a tour record was set aside. Song data came through Spotify's API, because its terms rule out running detection software on the service directly.

When the flagged artists were checked against Deezer's AI labelling, only 14% were confirmed as AI, and the blog does not give a count. Music Ally, which covered the study on 16 October, reports nine confirmed artists, none of whom mentioned AI in their profiles, all of them reached through Release Radar. The two outlets also differ on the listening window, 28 days in the blog and 25 in Music Ally, and Music Ally says the report's date is unclear.

The authors want mandatory, visible AI disclosure on track pages and recommendation playlists, listener controls to filter or reduce AI music, standard disclosure metadata from distributors, and regular public figures from platforms on how much AI music they recommend. They cite surveys in which 80% of listeners want AI music clearly labelled and 73% want to know when a platform recommends it.

Spotify added a "Verified" badge for human artists in May and an "AI Persona" badge in September, and says it will not rely on self-disclosure alone for the latter. Music Ally adds that the company says it is adjusting its algorithm to keep AI "slop" out of recommendations, Release Radar included. No date is given for further changes.

Comments