Recommendation SystemConsumer FacingVerified

Everything a user sees on Spotify's home screen — playlists, podcasts, albums, shortcuts — is ranked and ordered by AI based on that individual's listening patterns and predicted preferences.

Details

Uses a two-stage pipeline (candidate generation + ranking) and a multi-armed bandit framework balancing familiar and new content. Runs on TFX, Kubeflow, and Google Cloud. Models retrain weekly. Led by Tony Jebara (formerly Netflix). The system optimizes for long-term satisfaction, not just clicks.

Products affected

Spotify (Free and Premium)

Sources & Evidence

Company Disclosure

Company DisclosureVerified

The Rise (and Lessons Learned) of ML Models to Personalize Content on Home (Part I)

Spotify Engineering·Nov 2021

https://engineering.atspotify.com/2021/11/the-rise-and-lessons-learned-of-ml-models-to-personalize-content-on-home-part-i

Company DisclosureVerified

For Your Ears Only: Personalizing Spotify Home with Machine Learning

Spotify Engineering·Jan 2020

https://engineering.atspotify.com/2020/01/for-your-ears-only-personalizing-spotify-home-with-machine-learning

Other practices by Spotify

OtherSpotify has been accused of filling its own playlists with music from pseudonymous "ghost artists" — tracks commissioned at lower royalty rates — to reduce the royalties it pays out on popular ambient and focus playlists.OtherIn October 2025, Spotify announced a formal partnership with all three major record labels (Universal, Sony, Warner) plus Merlin and Believe to develop AI music tools that compensate artists — and disclosed it is building its own generative AI research lab.OtherSpotify uses AI tools to help screen job applications and match current employees with internal opportunities, though it discloses that AI assists rather than replaces human hiring decisions.

Have evidence about Spotify's AI practices? Submit a report.

Report a Sighting →