Checking a song you found
Use a file you can trace to the release you heard. Compare the result with artist credits, release notes and any production disclosure. A streaming preview or repost may cover only part of the recording you want to check.
Music origin check
AI music checker
Is this song AI-generated? Check an audio file for AI signals, then review what the result can tell you.
Check your audio
MP3 or WAV · Up to 20 MB and 15 minutes
01Choose audio02Run the check03Review results
Drop a song here
MP3 or WAV · Up to 20 MB and 15 minutesChecking service availability…
Check a song in three steps
1
Choose audio
Choose the MP3 or WAV you want to check. Preview it to make sure you have the right audio.
2
Run the check
Read the file-handling details, then submit. Keep the task open while the detector processes your audio.
3
Review results
Review the likelihood and analyzed range. Compare the result with the recording or production history.
How to read your result
Read the score together with the analyzed range and any separate track results. Each answers a different question about the audio you submitted.
Example values for this guide only. No audio was analyzed for this illustration.
Original audioAI likelihood
76%
Analyzed range0:30–1:30
Example recording length: 3:00. Highlighted range: 0:30–1:30.
Separate track estimates
- Vocals
- 18%
- Accompaniment
- 91%
- What the AI likelihood means
- A higher score means stronger AI signals to the detector. In this example, 76% is a model estimate for the analyzed audio. It does not mean that AI composed 76% of the song.
- Which part was analyzed
- Check the start and end times. Here, the estimate covers 0:30–1:30 of a three-minute recording. It cannot describe an intro, verse or outro outside that range. Review segment results when available.
- Why vocals and accompaniment can differ
- Original audio is the combined recording. Separate vocal and accompaniment estimates may disagree, especially in a mixed production. Read each on its own; averaging them does not measure the human contribution.
A missing score stays unknown
A dash means no usable value was returned. “No vocals detected” concerns the vocal analysis only. Neither that label nor an unsuccessful check establishes that a recording is human-made.
Put the result in context
An AI music check is a starting point for a closer review. The next step depends on what you need to understand.
Reviewing a playlist submission
Use the report to identify what needs a follow-up question. Ask the creator for the original export and relevant production notes, and compare them with the analyzed range. Give them a chance to explain an unexpected result.
Responding to an unexpected score
Keep original takes, separate tracks, your session project and dated exports together. Check that the report refers to the same file and version. If you dispute a result, share the relevant interval and the production evidence that explains it.
Prepare your audio for a useful check
Keep a clear link between the file you submit and the recording you want to understand.
Start with the original export
Choose the version you actually want to review and keep an untouched copy. A speaker recording or an edited repost introduces changes that can affect analysis. Preview the selected file before submitting it.
Check the format, size and length
Upload one MP3 or WAV at a time, up to 20 MB and 15 minutes. Renaming a file extension does not convert the audio. If needed, export a playable MP3 or WAV from your audio editor and listen to it first.
Document an excerpt
For a long recording, choose the passage you need to inspect and note its position in the original. Keep the source filename and version with the result. An excerpt provides information about that passage, not the rest of the recording.
How this AI music checker works
Audio is analyzed by ACRCloud. This page presents the returned estimates and their scope so you can review them alongside the recording’s history.
- What appears in your report
- The report shows the returned classification, AI likelihood and available analyzed ranges. Separate track results and a detector version appear when supplied. This page does not identify which music generator made a song.
- What can affect the estimate
- Audio quality, compression, mixing and editing can affect detection. ACRCloud’s published performance figures describe its own evaluation data; they do not establish accuracy for your recording. Keep source audio and production evidence with any result you rely on.
- Before sending your audio
- Submitting a check sends your audio to ACRCloud for analysis. Review the file-handling notice in the upload panel before sending unreleased or sensitive material. It provides the current handling details for this service.
Questions about AI music detection
No. A detector’s classification can be wrong. A high score is a reason to look closer; a low score does not prove an entirely human production. Check the source and production records as well.
A detector can classify an original recording incorrectly. Audio quality, mixing, compression and edits can affect the result. Preserve the original audio and compare it with production records.
When the detector returns separate vocal and accompaniment analyses, this page shows them separately. Missing stem results stay missing. The percentages are not a calculation of human creative contribution.
This page accepts audio files. A streaming link, preview or unidentified long mix is a different input. Do not treat a short preview’s score as a result for the complete song.
AI-origin detection does not determine copyright ownership, similarity to an existing song, permission to distribute, or whether a submission meets a platform’s rules.
The page keeps unavailable scores and unsuccessful checks separate from human and AI classifications. If a submission is still uncertain, check its status before starting another one.