Data AnalysisInternal OnlyVerified

Every song on Spotify is analyzed by machine learning to measure characteristics like danceability, energy, mood, tempo, and acoustics — data that powers all of Spotify's recommendation and discovery features.

Details

Traces to Spotify's $100M acquisition of The Echo Nest in 2014. Exposed through the Spotify Web API (though as of May 2023, developers are prohibited from using API data to train AI models). Internally powers "Truffle Pig," an editor tool that lets Spotify's curators search the catalog using feelings and adjectives. Glenn McDonald's "Every Noise at Once" map identifies 4,341+ genres from this data.

Products affected

Internal Editorial Tools

Sources & Evidence

Company Disclosure

Company DisclosureVerified

Get Track's Audio Features

Web API·Jan 2024

https://developer.spotify.com/documentation/web-api/reference/get-audio-features

Company DisclosureVerified

The Winding Road to Better Machine Learning Infrastructure Through Tensorflow Extended and Kubeflow

Spotify Engineering·Dec 2019

https://engineering.atspotify.com/2019/12/the-winding-road-to-better-machine-learning-infrastructure-through-tensorflow-extended-and-kubeflow

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 →