Blush tool detects Suno AI music via 60ms artifact
A new detection tool called Blush identifies Suno-generated tracks by measuring a residual autocorrelation spike, according to gamebusiness.jp.
An author has released Blush, a tool that detects music generated by Suno by measuring a residual autocorrelation spike at about 60 milliseconds, according to gamebusiness.jp.
The 60ms and 120ms spikes appeared in 36 of 40 Suno tracks tested, with median z-values of 12 and 14. None of 42 human recordings and none of 53 Udio tracks from 2024 showed the same spikes. The same pattern also appeared in another generator, YuE2.
The artifact shows up not only in raw Suno downloads but also in mixed tracks, allowing per-segment identification. Blush analyzes songs across seven layers, including high-frequency cutoff, noise floor, modulation, stereo correlation, repetition and residual autocorrelation.
The tool was made by an author who has produced music with Suno for nearly three years.
Deezer has operated its own detector since January 2025 and reported in that month that about 10% of daily new tracks, roughly 10,000 songs, were fully AI-generated. Deezer said the share rose to 39% (60,000 tracks/day) in January 2026, 44% (75,000 tracks) in April 2026, and above 50% at a monthly average of 90,000 tracks in June 2026.
Quick answers
What is Blush?
Blush is a detection tool that identifies Suno-generated music by measuring a residual autocorrelation spike at about 60 milliseconds.
How accurate is Blush at detecting Suno tracks?
The 60ms and 120ms spikes appeared in 36 of 40 Suno tracks tested. None of 42 human recordings and none of 53 Udio tracks from 2024 showed the same spikes.
Does Blush work on mixed tracks?
The artifact shows up in mixed tracks as well as raw Suno downloads, allowing per-segment identification.