How Spotify Uses AI to Know Your Next Favourite Song — Explained for Kids

You open Spotify. You hit play on a playlist. Song after song, it just gets you. How does a computer know you are in the mood for that exact song right now? The answer is one of the most impressive AI systems in the world.

Three AI Systems Working Together

Spotify does not use just one AI — it combines three different approaches that work together to understand your taste with uncanny accuracy.

1. Collaborative Filtering: Birds of a Feather

The idea is simple: if you and another listener both love Billie Eilish, Arctic Monkeys, and Clairo, you probably have similar taste. So if they discovered a new artist you have not heard yet, there is a good chance you will love them too.

Spotify does this at a scale of 600 million users [Spotify Newsroom]. Its AI maps everyone's listening patterns and finds your musical "neighbours" — people whose taste overlaps with yours. Then it surfaces music they love that you have not encountered yet.

2. Audio Analysis: Reading the Music Itself

Spotify's AI does not just track what songs people listen to — it analyses the actual audio. Every track is measured for tempo (how fast), key (what notes), energy (loud or quiet), danceability, acousticness, valence (happy or sad), and dozens of other features. This is how Spotify knows you like "high energy, upbeat, guitar-driven tracks" even when no human has labelled them that way.

3. Natural Language Processing: Reading About Music

Spotify's AI constantly scans the internet — blog posts, reviews, playlists, social media — reading how people describe music. When a music blog calls an artist "dreamy indie-pop with lo-fi vibes," the AI learns those associations and uses them to connect that artist to other music described in similar language.

Discover Weekly: The Monday Magic

Every Monday, Spotify's AI generates a fresh 30-song playlist for each of its 600 million users. That is 18 billion personalised recommendations, generated overnight [Spotify Engineering Blog]. Each one is different. Discover Weekly has introduced more people to their new favourite artists than almost any other music discovery tool in history.

What This Means for Artists

For musicians, Spotify's AI is both an opportunity and a challenge. A small independent artist with 500 followers can suddenly land in Discover Weekly playlists and wake up with 500,000 streams. The AI does not care about fame — it cares about fit. That changes the music industry in ways we are still figuring out.

3. Natural Language Processing: Reading What the World Says

Spotify's AI also reads text from across the internet — blogs, articles, and playlist names — to understand how people describe music. If thousands of people call an artist "chill study music," the AI learns to suggest that artist when you're in a focused mood. It's reading human words to understand sound.

Try This

Look at your "Discover Weekly" or "Made For You" playlist. Try to guess which of the three systems suggested each song: is it because similar listeners loved it (collaborative filtering), because it sounds like your favourites (audio analysis), or because of how it's described online (NLP)? You're now thinking like a recommendation engineer.

Frequently Asked Questions