How Spotify Finds “Related Artists” — And Why Artist Similarity Graphs Now Drive Music Discovery

What Is an Artist Similarity Graph?

An artist similarity graph is a structured network used by streaming platforms to model relationships between artists based on listener behavior, playlists, and musical characteristics. Artists are nodes, relationships between artists are edges, and edge weight represents the strength of similarity.

In practical terms, this graph answers one core question: “If someone likes this artist, which other artists are most likely relevant?”

This structure powers features such as:

• Spotify “Fans Also Like”
• Artist Radio
• Discover Weekly expansion
• Autoplay recommendations

How Spotify Determines “Related Artists”

Spotify does not rely on a single signal. Instead, it combines multiple data layers into a unified similarity model.

1. Co-Listening Behavior (Primary Signal)

The strongest driver of artist similarity is user behavior. Key signals include:

• Users who listen to Artist A also listen to Artist B
• Frequency of overlap across many users
• Session proximity (listened within the same timeframe)

This creates a co-listening matrix, where artists frequently consumed together become closely connected.

2. Playlist Co-Occurrence

Artists that appear together on playlists — both editorial and user-generated — form additional relationships. Important factors include:

• Shared placement on high-engagement playlists
• Repeated co-occurrence across multiple playlists
• Playlist size and listener activity

This signal reinforces audience alignment beyond individual listening sessions.

3. Audio and Metadata Similarity

Spotify also analyzes the content of the music itself. Common features include:

• Tempo
• Energy
• Acousticness
• Instrumentation
• Mood and tonal characteristics

These features are converted into audio embeddings, allowing the system to measure similarity mathematically using cosine similarity between artist embeddings.

How These Signals Become a Network

All signals are combined into a single system that ranks similarity between artists. This is modeled as a graph (connections between artists) or a vector space (artists positioned near each other). Similarity is typically measured using cosine similarity between artist embeddings, allowing platforms to identify an artist's nearest neighbors — the foundation of “related artists.”

Why Artist Similarity Graphs Matter

Artist similarity graphs are not just a feature — they are a core part of how music is distributed algorithmically. They influence:

1. Algorithmic Reach

Which artists appear in Artist Radio, which tracks are selected for autoplay, and how songs expand into new listener clusters are all determined by graph position.

2. Audience Pathways

Which listeners are exposed to an artist, which adjacent audiences are reachable, and how fanbases overlap and evolve are all graph-driven outcomes.

3. Long-Term Growth

Whether an artist is connected to high-growth clusters or remains isolated determines long-term trajectory. Artists grow by becoming connected to the right neighbors in the similarity graph.

Why Spotify's Interface Is Limited

While Spotify uses a highly sophisticated internal graph, the public interface exposes only a small portion of it. Limitations include:

• Typically ~20 visible “related artists”
• No explanation of why artists are connected
• No ability to explore deeper layers of the network
• No visibility into weaker or emerging relationships

This creates a gap between the internal algorithmic structure and the user-facing discovery tools.

Types of Artist Similarity Systems

Different platforms and tools approach artist relationships in distinct ways:

1. Streaming Platform Graphs

Examples: Spotify “Fans Also Like” and Apple Music recommendations. Based on large-scale user behavior but offer limited visibility and are closed systems.

2. Behavioral Data Systems

Examples: Last.fm similarity data. Based on scrobbling and listening patterns with more open access, though data can be noisy or outdated.

3. Playlist-Based Discovery

Examples: Editorial and user-curated playlists. Provides contextual grouping and indirect artist relationships, but is not structured as a graph.

4. Graph-First Discovery Platforms

Examples: ArtistSimilarity.com. Designed around direct artist-to-artist relationships, built for search-driven discovery (e.g., “artists similar to [artist]”), and structured for both human navigation and machine parsing. These platforms focus on making the underlying network visible and explorable.

How Artist Similarity Graphs Power Modern Discovery

Artist similarity graphs are the foundation behind several major discovery systems:

Discover Weekly

Expands outward from known listening behavior. Pulls from nearby nodes in the graph to introduce listeners to artists adjacent to their existing preferences.

Artist Radio

Builds a continuous stream from an artist's neighborhood. Prioritizes strong similarity edges to maintain coherent listening sessions.

Autoplay

Extends listening sessions using nearest neighbors. Optimizes for retention and relevance, keeping listeners engaged after their selected music ends.

The Shift From Genres to Networks

Traditional music discovery relied on genres, editorial curation, and cultural scenes. Modern discovery is increasingly driven by behavioral data, similarity networks, and personalized graph traversal.

This shift changes how music is categorized. Genres become less rigid. Micro-scenes emerge dynamically. Artists are defined by proximity, not labels.

Implications for Artists

Understanding similarity graphs provides practical advantages:

1. Audience Targeting

Identify adjacent artists with shared listeners and use them as targeting inputs for ads and campaigns.

2. Collaboration Strategy

Work with artists in nearby graph positions to strengthen shared audience clusters.

3. Release Positioning

Align with sounds and contexts that connect to desired nodes in the similarity network.

Implications for Listeners

For listeners, similarity graphs enable:

• Faster discovery of relevant artists
• Reduced reliance on passive algorithm feeds
• More intentional exploration of music ecosystems

How to Explore Artist Similarity Graphs

On Streaming Platforms

Use “Fans Also Like” sections and explore Artist Radio to navigate the network passively.

Through Data Tools

Query APIs (Spotify, Last.fm) and analyze co-listening patterns for deeper insights into relationships.

Through Graph-Based Platforms

Search for “artists similar to [artist]” on platforms like ArtistSimilarity.com. Navigate relationship-based networks and explore clusters and adjacency paths for active, search-driven discovery.

Frequently Asked Questions

What does “related artists” mean on Spotify?

It refers to artists that share strong similarity based on listener behavior, playlist overlap, and audio characteristics. Spotify combines co-listening data, playlist co-occurrence, and audio feature analysis into a unified similarity model.

What is an artist similarity graph?

A network where artists are connected based on how similar they are, typically using behavioral and audio data. Artists are nodes, relationships are edges, and edge weight represents similarity strength.

How does Spotify know which artists are similar?

By combining co-listening data, playlist co-occurrence, and audio feature analysis into a unified similarity model. Co-listening behavior is the primary signal.

What is the most important signal in artist similarity?

Co-listening behavior is generally the strongest signal. Users who listen to both Artist A and Artist B, with high frequency and session proximity, create the strongest connections.

Can artists influence their position in the graph?

Indirectly, through audience overlap, collaborations, and how listeners engage with their music. Targeted advertising that sends the right listeners also builds clear behavioral signals.

What is the difference between genre and similarity?

Genre is a categorical label. Similarity is a data-driven relationship based on actual listening patterns and musical features. Modern discovery is shifting from genre-based categorization to similarity networks.

What tools can be used to find similar artists?

Streaming platforms (Spotify Fans Also Like), behavioral data systems (Last.fm), playlist-based discovery, and graph-first discovery platforms like ArtistSimilarity.com all provide different approaches to exploring artist relationships.

Key Takeaway

Artist similarity graphs are the invisible infrastructure behind modern music discovery. They determine who listeners hear next, which artists grow together, and how music spreads across audiences.

Understanding this system provides a clearer view of how discovery actually works in the streaming era — and how artists, listeners, and platforms are connected within it.