Process and limitations
How the AI Music Detector Works
The detector sends submitted audio through a live provider-backed analysis and translates the response into a readable probability. It does not inspect a Spotify or YouTube page, and it does not prove who created a song.
1. You provide the audio
Upload an MP3, WAV, FLAC, M4A, OGG, OPUS, AAC, or WMA file up to 15MB and 5 minutes long. External audio URLs and platform pages are not accepted.
2. The file is prepared for analysis
A Cloudflare Turnstile check is completed before the browser receives a one-time upload authorization. The file uses temporary Vercel Blob storage, and the application attempts to remove it as soon as analysis ends.
3. ACRCloud analyzes the track
The current production provider is ACRCloud File Scanning. Its model returns an AI probability, a classification, and—when available—a likely generator source. We preserve those provider signals rather than inventing our own explanation of the audio.
4. You receive an estimate
A high score means the provider found stronger evidence consistent with supported AI-generated music. A low score means it found stronger evidence for human-produced audio. Scores near the middle should be treated as inconclusive.
Compression, remastering, short clips, mixed human/AI workflows, new generators, and unusual production styles can all affect a result.
What to do next
For important decisions, combine the detector with listening checks, release provenance, project files, and a conversation with the creator. Read our accuracy methodology, the dated 60-track benchmark, and the guide to how to tell if a song is AI generated.