How Classification Works

Traxsource Help Center

How is a track classified?

Every track is scanned to inform its label, and that analysis is combined with human review. Classification is per track, so each track is assessed on its own. We know AI detection is not a perfect science, which is why no label rests on a single automated check.

Who are your detection partners?

We work with two independent technology partners to inform our classification decisions.

SoundPatrol combines AI detection, neural fingerprinting, and large-scale audio analysis to help identify synthetic content and tell human-created and AI-generated music apart. Their technology looks at patterns, artifacts, and model-specific signatures within a recording.

SH Labs develops AI music detection technology focused on identifying characteristics commonly associated with generative audio systems. Their analysis examines spectral, harmonic, and temporal patterns within a recording, adding an independent layer of verification.

Is the detection perfect, and what if a label looks wrong?

No detection is perfect, and we know it. We recognize that AI detection is not a perfect science, so we have developed a straightforward review process and we are committed to getting it right.