What does 12% mean?
12% is the ensemble's combined 0–100 AI-likeness score. It is not detector accuracy, the song's AI share, or a calibrated 12% chance of AI authorship. Higher scores mean stronger AI-associated patterns.
AI MUSIC CHECK
AI Song Checker: See how likely a song is AI-generated
Use this AI song checker or paste a public link to see the AI-likeness score and model evidence.
Check one song
How to Use the AI Song Checker
Use an audio file or a public track link from Spotify, SoundCloud, YouTube, or Suno. This AI song checker reviews one song and shows its model evidence.
- 01
Choose one song
Upload an MP3, WAV, FLAC, OGG, or M4A file up to 80 MB and five minutes. Use the original mix when possible.
- 02
Add a public link
Paste a public single-track link. Private tracks, playlists, or region-restricted pages may not be accessible to the detector.
- 03
Read the evidence
The AI song checker shows the combined AI-likeness score, each model score, the score spread, and warnings. Read them together; no category proves authorship.
HOW TO READ YOUR RESULT
How to Read an AI Song Detector Result
The AI song detector combines one overall score with each model's output. Read them together to see which AI-associated patterns were found and whether the models agree. The estimate covers only the audio submitted.
What does 0–23% mean?
It spans the lowest and highest model scores. Here the models returned 0% and 23%; their equal-weight average is 11.5%, rounded to 12%. The range is not a confidence interval.
What does the verdict mean?
“More likely human-made” means the score falls in the detector's HUMAN category. It describes this model result; it does not verify who created the song.
Two models review different audio patterns
The current AI song detector combines two models equally. SpecTTTra checks patterns across a two-minute window; Fakeprint looks for regular spectral peaks. Scores differ when models find different evidence.
Read the detector model notesSpecTTTra α · 120 sec
Looks at patterns across a two-minute music window to assess song structure.
Fakeprint
Looks for regularly spaced spectral peaks associated with some neural audio vocoders.
FILES AND PUBLIC LINKS
What Can This AI Music Checker Analyze?
Check one audio file or public single-track link per request. This AI music checker accepts Spotify, SoundCloud, YouTube, Suno, and direct HTTPS audio when accessible to the detector.
Audio files
Upload MP3, WAV, FLAC, OGG, or M4A audio. Files can be up to 80 MB and five minutes long. A clean, higher-quality original gives the models more detail to analyze.
Public track links
The AI music checker can read public single tracks or videos from Spotify, SoundCloud, YouTube, or Suno. Playlists, private uploads, and inaccessible links may not work.
USE THE SCORE IN CONTEXT
What a Song Checker Can—and Cannot—Tell You
A song checker can help review a submission or your own AI-assisted production, but cannot identify a creator, establish copyright ownership, or prove misconduct. Newer generators may be missed; short clips, compression, remastering, pitch correction, and mixed human-AI production can change the evidence. Use the score to guide review, not to accuse an artist.
AI SONG CHECKER FAQ
AI Song Checker Questions and Answers
Learn what the scores mean, how the AI song detector compares model outputs, and when to treat a result with extra caution.
What does an AI song checker score mean?
It is an AI-likeness score from 0 to 100, not detector accuracy, the track's AI share, or a calibrated probability of authorship. Use an AI song checker as an initial signal, then consider the audio's source and production context.
How does an AI song detector combine its scores?
The current default ensemble combines two model outputs with equal weights. For example, scores of 23% and 0% average to 11.5%, which the page rounds to 12%. The overall score is still a model estimate, not a calibrated probability.
Why can an AI song detector show different model scores?
An AI song detector may return different scores because its models analyze different audio patterns. Compare both outputs and the score spread instead of relying on one number.
Can an AI music checker prove who made a song?
No. An AI music checker can provide a signal for further review, but it cannot verify authorship, copyright ownership, or misconduct. Do not use a score alone to accuse an artist.
What does a song checker’s 0–23% range mean?
It is the span from the lowest to highest model score. A wider range means more disagreement between models. It is not a confidence interval and does not say how likely the result is to be correct.
Which audio files and links can I check?
Upload MP3, WAV, FLAC, OGG, or M4A audio up to 80 MB and five minutes, or paste a public single-track Spotify, SoundCloud, YouTube, Suno, or direct HTTPS audio link.
Is the audio saved after a check?
Audio and links are processed for the requested check. Temporary audio files are removed after processing, and SunoPull does not keep a song-check history.
The selected audio or link is sent to the detector for this request. Temporary audio files are removed after processing; SunoPull does not create a saved result history.