Everything you need to know about AI music detection, how AISongScan works, and what your results mean.
When you upload a track, our system analyzes the audio at a deep acoustic level -- examining patterns in timing, pitch variation, texture, and structure that differ between human performances and AI-generated audio. The fingerprint is compared against known signatures of AI tools and a confidence score is returned.
AISongScan is trained to detect output from Suno, Udio, Sonauto, Mureka, and Riffusion. Detection coverage expands as new generation tools gain adoption. ElevenLabs, MiniMax, and Seed-Music detection is in development.
The underlying detection model reports 99.98% precision and 99.94% recall with a 0.06% false positive rate in its own evaluation. Those are the model vendor's laboratory figures, not measurements independently verified by AISongScan, and real-world accuracy varies with audio quality and production. Results are probabilistic -- a confidence score, not a legal verdict. No detector is 100% accurate and results should not serve as the sole basis for enforcement or legal decisions.
A score above 50% indicates the track likely contains AI-generated audio. A score below 50% suggests human-made audio. The closer to 100%, the stronger the match to known AI-generated audio patterns. A low score does not guarantee the track is fully human-made.
Heavy post-processing, mixing, or mastering can affect detection accuracy. For best results, upload original or minimally processed files. Intentional manipulation to evade detection is outside the intended use case.
Yes. Paste a Spotify, Apple Music, YouTube or Deezer link, or a direct audio file URL, and AISongScan will scan it. This is the quickest way to check released music, which is what usually arrives as a playlist pitch, a demo, or a sync submission.
Not quite. With a streaming link we identify the recording from the page's public title information and analyse that release, so the result tells you which track it matched. Uploading the audio file is the most accurate method because it analyses your exact recording. Use upload for your own, unreleased, or very new tracks.
Not every release is in the analysis catalogue. Coverage varies by release and is not predictable by age or label. When a link cannot be matched, upload the audio file instead. A scan that fails this way does not count against your allowance.
MP3, WAV, M4A, and FLAC are all supported. Upload the original file before heavy compression for the most accurate results.
Yes. Individual scans are limited to standard file sizes appropriate for full-length tracks. Very large files may need to be trimmed before upload.
No. Your audio is sent over an encrypted connection, analyzed in transit, and never stored on AISongScan servers after the scan completes.
No. Your audio is never used to train AI models, and it is never sold or shared for marketing. To produce your result, audio is processed by a specialist audio-recognition provider acting on our behalf, as described in our Privacy Policy.
You can request data deletion from within the app. For account closure, contact contact@stablestateit.com.
There is a free tier with a limited number of scans. Creator ($9/month or $90/year) and Professional ($29/month or $290/year) plans provide higher scan limits and additional features.
Billing is handled securely via Stripe. You can manage your subscription from within the app. Cancellations take effect at the end of the current billing period.
An account is required. Sign in uses a passwordless magic link -- no password to remember or manage.
Upload any audio file and get your AI detection result in seconds.
Start Scanning Free