We analyzed the top 20 songs on Spotify's Global chart, separating each track into its component parts to detect AI-generated content at the stem level. Our findings reveal a nuanced picture of how AI is being used in today's most popular music.
Most AI detectors analyze the full song as one audio file. But that misses a lot. A track might have human vocals layered over AI-generated beats, or mix live guitar with synthesized bass. So we break each song down into its individual parts and check each one separately.
We run each track through a neural network trained on thousands of songs to isolate the individual stems. The model outputs four primary stems (vocals, drums, bass, and "other"), then we further process the "other" stem using spectral analysis to extract melody and harmony components.
Lead vocals, backing vocals, and any human voice elements
Percussion, drum machines, and rhythmic elements
Bass guitar, synth bass, and low-frequency elements
Lead synths, guitar riffs, and melodic instrumental lines
Chords, pads, and harmonic accompaniment
Each stem gets uploaded to our detection service, which compares audio fingerprints against databases of known AI-generated music and analyzes spectral characteristics that differ between human and AI compositions. The service returns a probability score—anything above 50% we flag as AI-detected.
If any of the five stems scores above 50%, we mark the song as "contains AI." This means a track with human vocals but AI-generated drums still gets flagged. We show which specific components triggered the detection so you can see exactly where the AI was used.
Here's what we found when we analyzed the top chart songs through our component-level detection pipeline.
Loading analysis data...
The data shows AI is already in chart-topping music, but it's not what you'd expect. We're not seeing fully AI-generated songs. It's more subtle—an AI drum loop here, a synthesized bass line there. AI as a production tool, mixed in with human work.
Listeners should know when AI was used to create the music they're streaming. Right now, most don't.
If a beat was made by AI, should the producer still get full royalties? These questions need answers.
AI can clone vocal styles and production techniques. Artists need tools to detect when their sound is being replicated.
Should AI-generated content get royalties? If a producer uses AI for the drums, do session drummers lose work? How much AI is too much before you have to disclose it? No one knows. The industry is still figuring it out.
These aren't hypotheticals. Our data shows AI is already in the charts. The rules are being written right now.
We're not anti-AI. We just think people should know what they're listening to. Labels, publishers, and streaming platforms need this information to make informed decisions about royalties and credits.
We'll keep running these analyses and publishing what we find.
Our component-level AI detection is available for artists, labels, and publishers.
Get Early Access