← Back home

How to read AI roles & scores

TRAICE shows (1) what AI did, and (2) how consistent + clear the disclosure is.

AI roles

“Roles” describe *where* AI was used in the creative process (e.g., vocals, composition, mixing, cover art). A track can have multiple roles, each with its own tool + description.

Extent labels

  • Documented only: AI wasn’t used for that role (0%).
  • Light assist: idea starter / reference / small edits.
  • Co-creator: meaningful contribution alongside humans.
  • Mostly AI: AI did most of the work for that role.
  • Fully AI: AI generated the role output end-to-end.

Scores (what they mean)

Important: TRAICE scores are not a “morality” or “integrity” rating.

Scores only measure how clearly and consistently your disclosure is documented (tools + roles + descriptions + your AI% claim lining up). A low score usually means “something doesn’t match” — not “you did something wrong.”

Transparency Score

A blended score of disclosure quality: completeness + consistency + clarity. If you picked a scenario/AI% that doesn’t match the roles you disclosed, transparency can drop.

AI–Human Spectrum (%)

An estimated AI influence percentage inferred from your AI roles + “extent” labels. It’s a heuristic (not a claim about authorship). Think: “how much of the final audio output was generated/processed by AI tools.”

Spectrum Alignment

Checks whether your claimed AI% is consistent with what the role matrix implies. Big mismatches get flagged because they confuse listeners/platforms reading the disclosure.

Scenario Match

Scenarios are presets that suggest a typical AI% range (e.g., “AI vocals + AI instrumental” usually lands higher than “AI mastering only”). If your role matrix falls outside that range, scenario match may drop.

Source Risk

A separate indicator about tool provenance/training clarity. It does not mean your release is illegal — it just reflects how transparent the tool’s training/licensing is publicly.

If anything looks off, it usually means one of these is mismatched: roles, extent labels, AI% claim, or scenario. You can edit the disclosure anytime.