How Are Spotify Playlists Curated? | DJ Will Gil

Spotify’s playlist machine is the most influential music recommendation system in the world. With 713 million monthly listeners and nearly 9 billion playlists created across the platform, the answer to “how are Spotify playlists curated” is simultaneously simple and structurally complex: some are built by humans, some are generated by machine learning, and most are made by the listeners themselves. The interplay between those three sources is what determines which artists break through and which stay invisible.

This guide explains exactly how Spotify’s curation actually works in 2026: who the editorial curators are, what the algorithm is measuring, how the BaRT recommendation system decides which tracks reach which listeners, what the new Prompted Playlist feature changes, and what every artist and listener should understand about the way the platform turns 100 million-plus tracks into the personalized listening experience that arrives on the home screen each morning.

Watch DJ Will Gill perform live. Contact him now to book your event.

The Three Tiers of Spotify Playlist Curation

Spotify’s playlist system is not a single recommendation engine it is three different curation models running in parallel and feeding each other constantly. Understanding how each tier works individually is the prerequisite for understanding how the platform recommends music as a whole.

According to NoMusica’s 2026 breakdown of the Spotify playlist ecosystem, the three tiers serve fundamentally different functions. Editorial playlists are built by Spotify’s internal team of music experts. Algorithmic playlists are generated automatically by machine learning models trained on listener behavior. User-generated and independent curator playlists are built by everyone else artists, brands, listeners, and tastemaker accounts that have grown audiences inside the platform. The boundaries between these tiers are blurring as Spotify’s recommendation system increasingly pulls signals from all three, but the tiers remain meaningfully distinct in how they are made, who controls them, and how artists access them.

How Spotify’s Editorial Team Actually Works

Spotify’s editorial team is a global group of full-time music experts based primarily in Stockholm, New York, London, and Los Angeles, organized by genre and region. Each editor is responsible for a specific portfolio of playlists, and editorial decisions are guided by both expertise and behavioral data from the platform itself.

The pitch process is the only legitimate way to be considered for editorial placement. Artists submit unreleased tracks through the Spotify for Artists tool, providing genre tags, mood descriptors, instrumentation details, and a written pitch describing the track’s context and audience. According to iMusician’s 2026 playlisting guide, pitches must be submitted at least seven days before the release date, the window closes the moment a track goes live, and there is no retroactive submission window. Editors review pitches against the playlist’s existing identity, the artist’s data profile, and the broader release calendar before making placement decisions.

The relationship between editorial and algorithmic curation has grown closer over time. Music industry analysis from Vohnic Music in 2026 notes that while Spotify maintains editorial playlists are curated by humans, the platform’s scale, machine learning infrastructure, and the data signals editors already rely on make it plausible that AI systems help filter and prioritize candidate songs before editors make final decisions. The company has not publicly confirmed algorithmic involvement in editorial selection, but editors are known to consult performance data such as save rates, skip rates, and streaming velocity all of which are generated by algorithmic systems when evaluating which songs earn placement on flagship playlists.

How the Algorithm Curates: BaRT and the Three Core Systems

Spotify’s algorithmic playlists Discover Weekly, Release Radar, Daily Mix, Spotify Radio, Autoplay, and On Repeat are generated by a machine learning system Spotify calls BaRT, short for Bandits for Recommendations as Treatments. According to Playlist Push’s algorithm documentation, BaRT analyzes listener behavior in real time and adjusts the user experience based on how each listener has reacted to music previously served to them. The “bandits” terminology refers to the multi-armed bandit problem in machine learning the algorithm continuously balances exploitation (serving songs it knows the listener likes) with exploration (testing new songs that might extend the listener’s taste).

BaRT is powered by three core analytical systems working in combination. Chartlex’s 2026 algorithm guide describes them as collaborative filtering, audio analysis, and natural language processing. Collaborative filtering identifies listeners with similar taste profiles and uses one listener’s behavior to predict another’s. If two users overlap heavily in their listening history, songs that one of them saves become candidates for the other’s Discover Weekly. Audio analysis runs every track through a system that evaluates tempo, key, energy, danceability, acoustic qualities, and dozens of other sonic characteristics, allowing the algorithm to match songs to listeners whose taste history aligns with those qualities. Natural language processing scans blog posts, reviews, playlist titles, and editorial content across the open web to understand how a track is being described culturally giving the algorithm context that pure audio data cannot provide.

The combination of those three signals is what makes Discover Weekly feel personal even though it was built by software. The algorithm is not just matching genres it is triangulating between behavioral data, sonic fingerprints, and the cultural language people use when they talk about music. According to Vohnic Music’s 2026 algorithmic placement analysis, the system produces meaningfully more long-term growth than editorial placements for most independent artists, because it continuously introduces music to new listeners based on engagement patterns, while editorial placements deliver concentrated short-term spikes.

The Signals That Trigger Algorithmic Placement

Understanding what the algorithm is actually measuring is the most useful answer to “how are Spotify playlists curated” for any artist trying to grow on the platform. The algorithm does not care about raw stream counts. It cares about whether listeners stay engaged with a track once they encounter it.

According to Chartlex’s analysis of more than 2,400 artist campaigns, the algorithm in 2026 weights save rate and repeat-listen ratio roughly 3x higher than raw stream volume when deciding which tracks to push into Discover Weekly and Release Radar. Tracks that maintain a save rate above 20% in their first two weeks are significantly more likely to enter algorithmic playlist rotation. Tracks that maintain a stream-to-listener ratio above 2.0 meaning each listener plays the song more than twice on average consistently trigger algorithmic placement within 10 to 14 days of release. The behavior signals the algorithm watches most closely, in roughly descending order of impact, are: saves and library adds, repeat listens, completion rate (how often listeners hear the song to the end), low skip rate (especially in the first 30 seconds), and additions to user-generated playlists. Strong performance on those signals during the first two weeks after release is what triggers the algorithmic snowball.

One important nuance from Chartlex’s 2026 algorithm analysis: in 2024 and 2025, Spotify shifted its algorithm to prioritize familiarity and retention over adventurous discovery. The system learned that playing songs listeners already know, or songs very similar to ones they know, keeps them on the platform longer. The practical implication is that breaking out of an existing audience requires more deliberate algorithmic teaching than it used to following new artists, exploring editorial playlists outside the listener’s usual lane, and intentionally searching for unfamiliar music are now the mechanisms that train the algorithm to expand a listener’s taste profile.

The 2026 Shift: Prompted Playlists and User-Steered Curation

The most significant change to Spotify’s curation system in years arrived in December 2025 with the launch of Prompted Playlist. According to Spotify’s official announcement from Co-President Gustav Söderström, the feature lets users describe in plain English exactly what they want to hear and instruct the algorithm to build a playlist that matches. The example Söderström gave: a user can prompt “songs from artists who are headlining major tours right now” and Spotify will generate a personalized playlist that draws on the listener’s complete listening history all the way back to day one filtered through that real-world prompt.

The strategic significance is that Prompted Playlist is the first time Spotify has handed direct algorithmic control to the listener. Previous personalization features used implicit signals what listeners played, what they skipped, what they added. Prompted Playlist accepts explicit signals stated in natural language. Söderström has positioned the launch as the beginning of “a new era of listener control,” and the company has acknowledged that nearly 9 billion playlists have been created by Spotify users to date proof, in their framing, that human curation remains the heartbeat of the platform even as machine learning becomes more sophisticated.

For artists, the implication is that algorithmic curation is shifting from a passive system that observes listener behavior to a directable system that listeners can command. The kinds of prompts users invent will become a new vocabulary the algorithm learns from, and the artists whose music maps cleanly to specific contexts (“songs for cooking dinner with friends,” “music that sounds like a cold morning in October”) may find themselves in entirely new pockets of discovery that the previous algorithm could not access.

What This Means for Listeners and Artists

For listeners, the practical answer is that Spotify’s recommendations are shaped just as much by what you do as by what Spotify decides. Saving a track teaches the algorithm something specific about your taste. Skipping the first 30 seconds of a song teaches it something different. Following a new artist explicitly tells the system to include them in your Release Radar. Searching for unfamiliar music outside your usual lane breaks the familiarity bias that has crept into the algorithm over the last two years. The home screen is not handed down it is co-authored.

For artists, the answer is that algorithmic playlist placement is earned, not pitched. Editorial playlists can be pitched through Spotify for Artists with at least seven days of lead time before release, and that pitch tool remains the only legitimate path to flagship editorial slots. But the more durable, longer-term growth almost always comes from algorithmic curation and that means designing every release for the engagement signals the algorithm cares about. A strong hook in the first 30 seconds reduces skip rate. A pre-save campaign produces immediate library adds on release day. Mobilizing existing fans in the first 24 hours seeds the algorithm with high-quality early behavior. According to MusicPromoToday’s 2026 promotion analysis, the platform’s detection systems for fake streams have become significantly more sophisticated, so attempts to game the algorithm with artificial activity now consistently produce track removal or account penalties rather than gains.

The honest summary of how Spotify playlists are curated in 2026: a global editorial team makes the most visible decisions, a machine learning system makes the most consequential ones, and the listeners themselves provide the data that powers both. The best way to navigate the platform whether as an artist trying to be heard or a listener trying to find music worth hearing is to understand which tier you are interacting with, what signals it responds to, and how to feed those signals deliberately.