You upload a track. A scan flags part of the audio. Does that mean the track was made with AI?

The useful next step is to examine what the scan actually found. TRACE reports measurements and provenance clues in the submitted file. Those findings help you decide where to investigate; they do not reconstruct every decision made in the studio.

Three different questions

A report brings together questions that sound similar but mean different things.

• What information is present in the file?
• What audio characteristics were flagged?
• Where do those characteristics appear across the track?

Available provenance can offer context about a file’s history. Audio measurements describe the sound being examined. A timeline shows the location of flagged measurements. Read these together rather than treating any single label as the whole story.

What “signals worth a closer look” means

This result identifies suspicious measurements without confirming AI generation. It gives you a place to start: inspect the highlighted region, listen to it in context, and compare the finding with what you know about the production.

A region might contain only a small part of a master. The report alone cannot tell you which instrument, plugin, sample or production step caused that measurement. A master combines those elements into one audio signal.

Evidence strength describes the strength of the reported measurements. Evidence coverage describes their reach across the track. Neither should be read as a guaranteed probability that an artist used AI.

No significant evidence is still a limited result

A scan with no significant AI signals means the submitted audio did not produce significant findings under that analysis. It is not a certificate of human authorship.

The same caution applies to provenance. Content Credentials provide a framework for recording and checking media history. Their presence or absence should not be treated as a universal judgment about a track. A normal export without such credentials can still have an ordinary production history.

A practical way to read your report

Start with the summary, then examine the highlighted regions. Listen before and after each region so that you hear its musical context. Keep your production notes, source sessions and export details alongside the scan if you need to explain how the track was made.

If you compare two files, record what they are. They might be two exports of one session or entirely different tracks. A difference in their reports is a difference in the submitted evidence, not an automatic explanation of what changed.

What this explainer establishes

This article explains how to interpret a screening report. It does not present a benchmark, a measured accuracy rate or a new detection experiment. Stronger claims would need documented test material, a defined method and a clear account of the limitations.

The useful question is not just “Was AI detected?” It is “What was found, where was it found, and what additional evidence would help me understand it?”