Guide / Field notes
How to Triage a Music Catalogue Without Automating Accusations
Design a queue, not a mass verdict.
· 12 min read
Input preparation
Define the business question, deduplicate files, preserve stable identifiers and use the best available masters. Record format and duration so later reviewers understand quality limitations.
Batch and triage
Run files through one documented version and sort results into routine review, elevated review, inconclusive and no-follow-up queues. Do not convert provider percentages into invented legal thresholds.
Apply base-rate discipline
When true incidence is low, even a useful classifier can produce more false flags than real finds. Sample every queue, watch clusters by genre and encoding, and estimate review burden before expanding.
Escalate and document
Ask for provenance on selected cases, give submitters a correction path and use neutral language such as 'requires review.' Preserve model version, date and final human disposition.
Key takeaway
The short version
Large-scale screening succeeds when it limits claims, samples its own errors and preserves a fair escalation record.
Run a free detectionFrequently asked questions
No. It can support a review, but origin and authorship require provenance and context. Do not use one automated or acoustic observation as proof.
Keep reading
Guide
How to Check a Song for AI: An Evidence-Led Workflow
A step-by-step method combining ears, measurements and provenance — in the order that actually works.
Comparison
Suno or Udio? Why the Production Workflow Matters More
Same goal, different machinery — and different tells left behind in the audio.
Industry
Does Spotify Detect AI Music? Separate Policy from Audio Analysis
Moderation, fraud checks and disclosure rules — a different job from acoustic detection.
Legal
AI Music Copyright: A Map of Clear and Unsettled Questions
Authorship, training data and voice likeness, sorted into settled and unresolved.