Free unlimited single-song checks · No signup · Results in seconds
AI Music Detector
Screen a song for signs of AI generation. Upload the audio or paste a public Suno song or direct audio link for a screening score — or an inconclusive result when the evidence is weak.
Drop an audio file here, or click to browse
MP3, WAV, M4A, FLAC, OGG or AAC · up to 50 MB
Free · No signup · Uploads deleted after analysis
Current evaluation: 68 determinate and 8 inconclusive results across 76 files — see the full per-track results.
Screening many tracks at once? Open batch screening in your dashboard →
What your report tells you
Very likely AI-generated
midnight-drive-final-master.mp3
Illustrative result: spectral fingerprint analysis. This example is not a measured sample or a generator attribution.
An uncalibrated screening signal — not proof of authorship or a judgment of the work's value.
- A score with limits. The acoustic score is an uncalibrated screening signal. Weak evidence receives an inconclusive result instead of a percentage verdict.
- Why a verdict was withheld. A short explanation when inconsistent or limited spectral evidence prevents a firm result.
- Declared provenance when present. Embedded Content Credentials appear separately. The acoustic model does not identify which generator made a song.
- A result you can share. Screenshot-friendly card with one-click copy for curators, labels and contest juries.
How the AI music detector works
Give it the track
Drop an MP3, WAV, M4A, FLAC or OGG file, or paste a link to the song. No account, no email.
We listen like a machine
The engine decodes the audio and scores spectral features from the first 300 seconds. File quality and signal confidence determine whether a verdict can be shown.
Read the verdict
Read the screening signal or the reason it is inconclusive, with an uncalibrated confidence indicator and a result you can copy.
What this AI music detector checks
Whether you call it an AI music checker or an AI music detector, the job is the same: recognize output from the major AI music generators — the tools behind most AI songs circulating today — and separate it from human-made recordings, including heavily-produced studio tracks. If a suspicious track smells like Suno output specifically, the dedicated Suno AI detector goes deeper; Udio-leaning tracks match the Udio AI detector; if your question is the plain “is this song AI?”, the Is This Song AI? checker answers exactly that. And if you're judging the voice itself — AI vocals inside a song, or an AI cover — the AI voice detector is the dedicated door. And if you'd rather judge a track with your own ears first, the how to tell if music is AI-generated guide walks all eight tells with a verification step for each.
Straight talk about accuracy
Full-length tracks from major generators are where detection is strongest. Short clips under ~15 seconds and heavily processed audio (aggressive mastering, vinyl rips, voice memos of speakers) are where any detector — including this one — is least certain, and we show reduced confidence instead of hiding it.
Both false positives and false negatives are possible. Compression can erase evidence, and agreement between model passes can still be wrong. “Inconclusive” means the file gives us too little evidence. Treat scores as screening signals; when money or reputation is on the line, pair the result with project files, stems or session data. The full detection methodology — signals, coverage, v6 testing — is documented.
Published results and limits
We publish scores, failures and withheld verdicts together. Our latest local replay uses the single-song policy and includes development data, source notes and file fingerprints. It is not a blind accuracy test. A separate 72-file generator challenge still produced 22 false-human results; those failures are published alongside the improvements.
76
Published test files
59 source files and 17 transformations, with per-file scores and source gaps disclosed. These are not 76 independent works.
68
Determinate results
No false positives or false negatives observed among these results. This development collection does not establish general accuracy.
8
Inconclusive results
Weak signals receive no firm verdict. Our research explains the compression failure we found, the correction and the cost in coverage.
Every number above traces to the published methodology — per-track results, fingerprints and provenance — read the full methodology.
Who it's for
Playlist curators
Filter AI spam out of the submission pile before it hits your playlist.
Labels & A&R
Know what you're signing. Flag undisclosed AI tracks in demos and catalogs.
Contest organizers
Enforce no-AI rules with a consistent, documented screening step.
Curious listeners
That viral track sounds too polished? Check it in seconds, for free.
Pricing
Single checks are free and unlimited. Paid packs buy the batch workflow — resumable multi-track screening, per-item retry and CSV export. No subscription, ever.
Free
$0
Unlimited single checks, no account needed.
Scan a track freeStarter
$4.90
60 batch credits, valid 12 months. One-time.
Get StarterPro
$9.90
240 batch credits, valid 12 months. One-time.
Get ProBatch screening runs in your dashboard workspace.
Multi-file upload, direct audio links, per-item retry, resumable batches and CSV export — one credit per successful track.
Open batch screening →Frequently asked questions
Is this AI music detector really free?
Yes — single track checks are free and stay free, no account needed, same engine and same scope for everyone. For screening many tracks at once, one-time batch credit packs (from $4.90) add resumable batches, per-item retry and CSV export — see Pricing. API access for labels and platforms is in evaluation; contact us.
Is this an AI music checker or an AI music detector?
Both are names for screening audio for signs of AI generation. We report a spectral score when the evidence supports a result and withhold a verdict when it is weak. The acoustic model does not identify a specific generator.
Does it work as a fully online AI music detector?
Yes — everything runs online in your browser: upload a file or paste a link on this page, no app, no extension, no install. Analysis takes seconds and the audio is permanently deleted right after.
Can I tell if a song is AI by ear before checking?
Often, yes — the common tells are vocals that never breathe wrong, glassy over-smooth production, lyrics that rhyme but don't mean much, template arrangements, and instrument details that fall apart under close listening. The full guide to how to tell if music is AI-generated walks all eight signs with a verification step for each; the checker then confirms with a score.
How accurate is the AI music detector?
Our October 10 v2.2 replay has 76 development files: 68 determinate and 8 inconclusive. All 25 AI files, including five heavily transcoded controls, now receive AI results. No determinate errors were observed in that collection. A separate 72-file AIME generator challenge exposed 22 false-human results and 38 inconclusive results. These include related transformations, not independent songs, and do not establish broad accuracy. We publish failures and sources alongside the improvements. Treat results as uncalibrated screening signals, not proof.
What files and links can I check?
Upload MP3, WAV, M4A, FLAC, OGG or AAC up to 50 MB, or paste a public Suno song share link or direct audio link. YouTube links are not supported; upload an audio file you have permission to use instead. Playlists, live streams, private sources and Spotify links are not supported. If a source cannot provide audio, upload the original file instead.
Do you store the songs I upload?
No. Audio is analyzed immediately and permanently deleted right after — including the transient copy in our transfer storage. We never build a library of your uploads. (If you use the optional batch workflow, each item likewise drops as soon as it finishes, with a 24-hour cap on stray transfer copies and 30-day metadata retention — see the privacy page.)
Do you use my uploads to train AI models?
No. Your audio is never used for training. Detection runs against the engine's existing reference fingerprints, and the file is discarded once your score is produced.
Is my audio shared with third parties?
No. Detection runs in-house on our own infrastructure with an open-source model — your audio never leaves this site, is analyzed once, and is deleted right after your score is produced. We don't sell data, and we don't run ad trackers.
Can it tell which AI model made a song?
The acoustic model does not identify a specific generator. Embedded Content Credentials may separately declare an origin such as Suno; that metadata signal is distinct from the acoustic assessment.
Can I use a result in a dispute or platform appeal?
A detection report is evidence-grade signal, not certification. Platforms and contest judges generally want corroboration — project files, stems or distribution records. Use our report as one input in that case.
Think you can tell AI from human by ear? Prove it.
Check a track nowSuno-specific checks: Suno AI Detector · Full technical detail: detection methodology