Spotify unveils Taste Profile – manual control over the AI algorithm

Spotify unveils Taste Profile – manual control over the AI algorithm


At SXSW, Spotify co-CEO Gustav Söderström announced a major shift in how the platform handles recommendations: the Taste Profile. This new AI-powered feature allows users to directly intervene in their discovery algorithms, offering a level of manual customisation previously unavailable on the platform. The feature is set to enter beta for Premium users in New Zealand in the coming weeks.

Read: Bandwidth Blog & Smile 90.4FM Tech Tuesday: Apple MacBook Neo!

The Taste Profile interface provides a detailed summary of a user’s current listening habits across music, audiobooks, and podcasts. At the heart of the feature is a “Tell us more” prompt, which serves as a direct line of communication with Spotify’s recommendation engine.

Users can use this tool to:

  • Prune their algorithm: Specifically request to remove genres or artists that are overrepresented in their feeds.
  • Contextualize listening: Provide ambiguous or lifestyle-based prompts. For instance, a user can inform the AI they are training for a marathon to receive more upbeat tempo music, or request news podcasts specifically for their morning commute.

For those who prefer a more passive experience, Spotify clarified that Taste Profile is entirely optional; users can ignore the feature and continue using the service as they always have.

See also

Taste Profile follows closely on the heels of Prompted Playlist, a feature released last month that allows users to generate specific music collections based on hyper-niche requests, such as “only songs featured in Stranger Things.”

By testing Taste Profile in New Zealand first—the same strategy used for Prompted Playlist—Spotify appears to be establishing a reliable pipeline for rolling out AI tools before expanding them to larger markets like the US and Canada. This latest update reinforces Spotify’s goal of moving away from a “black box” algorithm toward a more collaborative relationship between the listener and the AI.