Machine Learning Deep Dive

How the Spotify Algorithm Works in 2026: An Inside Look

Updated: February 2026 · 7 min read

Spotify uses one of the most sophisticated AI recommendation ecosystems in the world, known internally as BaRT (Bandits for Recommendations as Treatments).

How BaRT Decides What Songs to Push

BaRT constantly runs mini-experiments by placing unfamiliar songs in front of small test listener cohorts. If the test cohort exhibits high engagement (listens through, saves, adds to playlist), BaRT expands the song's radius to tens of thousands of similar listeners.

Three Signals That Win the Algorithm

• Low 30-Second Skip Rate: If a user skips your song before 30 seconds, BaRT treats it as a negative affinity signal.
• High Library Add Rate: Adding your song to a personal playlist or user library is the highest weighted organic engagement signal.
• Consistent Inflow: Steady daily stream velocity outperforms sporadic one-day spikes.