How Ghost Detect Works

What Are Ghost Artists?

Ghost artists are fake personas on streaming platforms. They don't exist as real musicians — there's no person behind the name, no live performances, no fan community. They're created to fill algorithmically-curated playlists with cheaply produced or AI-generated tracks, earning streaming revenue without the overhead of being an actual artist.

Ghost artists are a growing problem for music communities. When playlists get flooded with ghost-generated content, real independent artists lose visibility. Subreddits that promote new music become targets for ghost artist spam — automated posts pushing links to these fake profiles.

How Ghost Detect Works

Ghost Detect identifies ghost artists by analyzing their digital footprint across the internet. A real musician leaves traces everywhere — music databases, concert listings, fan wikis, review sites, social media, and collaborations with other artists. A ghost artist exists almost exclusively on the streaming platform.

The tool checks an artist's presence across multiple independent sources including MusicBrainz, Wikipedia, Last.fm, Genius, Discogs, Bandcamp, and setlist.fm. It looks at whether the artist has entries in music databases, whether they have a history of live performances, whether other verified artists have collaborated with them, and dozens of other signals that indicate a real person or group behind the music.

These signals are combined using a dual-pass machine learning system trained on over 2,000 verified artists — both confirmed humans and confirmed ghosts. The first pass analyzes the artist's external presence. The second pass examines their collaborator network, checking whether the artist has worked with other verified human artists. This two-stage approach delivers accurate results even for emerging artists with small online footprints.

Over 3,500 confirmed ghost artists from curated blocklists are recognized instantly with zero API calls or processing delay.

What Ghost Detect Does NOT Do

• It does not analyze the music itself (audio quality, composition patterns)
• It does not access any user's streaming account or listening data
• It does not make judgments about music quality — a real artist making simple music is still a real artist
• It does not store any personal information about users

Accuracy

The model correctly classifies over 99% of artists in cross-validated testing, with strong performance on both established and emerging artists. It's designed to be conservative — it would rather flag an artist as uncertain than incorrectly label a real musician as a ghost.

Some legitimate artists with very small online footprints may initially be flagged as uncertain. This is expected for brand-new or extremely underground artists who haven't yet built a presence outside of streaming platforms. The collaborator analysis pass often resolves these cases by detecting connections to verified human artists.

How to Use Ghost Detect

Open Source Data

Ghost Detect uses data from publicly available music databases including MusicBrainz, Wikipedia, Last.fm, Discogs, Genius, Bandcamp, and setlist.fm. No proprietary data sources are used.