Provider-backed audio analysis
AI Music Detector
Upload a song to estimate whether it was AI-generated. Get a probability score and the signals returned by the live detection provider.
Use the Detector in Your Language
English
You are using the English AI music detector.
How It Works
1. Provide audio
Upload a supported audio file up to 15MB and 60 seconds long.
2. Analyze the track
Our ACRCloud-backed workflow compares the audio with its current detection models.
3. Review the estimate
Use the probability and likely source as evidence, never as proof on their own.
AI Song Detector, Music Checker, and Second Opinion
“AI music detector,” “AI song detector,” and “AI music checker” all describe the same practical task: gathering evidence about a track's origin. The score is most useful alongside listening checks, source information, and project files.
Start with our listener-friendly guide to how to tell if a song is AI generated, then use the upload tool for a separate probability estimate.
Frequently Asked Questions
How can I tell if a song is AI-generated?
Listen for repeated phrasing, unstable vocals, unnatural transitions, and inconsistent room sound, then use a detector as a second signal. No single clue or probability score proves how a song was made.
What audio can I check?
Upload an MP3, WAV, FLAC, M4A, OGG, OPUS, AAC, or WMA file up to 15MB and 60 seconds long. Spotify, YouTube, SoundCloud, and external audio URLs are not supported.
Does the detector work for Suno and Udio tracks?
Our current provider returns source-level signals for supported AI generators. In our dated 60-track benchmark it classified all 20 sampled Suno tracks and all 20 sampled Udio tracks as AI-generated. Coverage can change as generation models evolve.
How accurate is the AI music detector?
Detection is probabilistic, not absolute. In our August 9, 2026 blind benchmark, the detector classified 60 of 60 sampled tracks as expected, including 20 human performances, 20 Suno tracks, and 20 Udio tracks. This batch result does not mean universal 100% accuracy.
See the 60-track benchmark and raw results →
See what happened after MP3 compression →