Distributors ask about AI use differently at the point of delivery

The audio task

At the point of delivering a finished release, some distributors now ask whether AI tools were used and, increasingly, what kind of AI tool. A producer who kept a provenance log during production answers that question directly; one who did not has to reconstruct the session's history under time pressure, right before release.

What the documents show

DistroKid's own help-centre article states that AI-made music is accepted if the uploader owns 100% of the necessary rights, does not impersonate another voice or likeness, is not 'mass-generated spam,' and does not infringe anyone else's rights; it does not ask, on this page, what specific AI tool was used or how its training data was sourced. TuneCore's GenAI Music Content Framework article takes a different approach, stating that 'TuneCore only distributes music created using GenAI models trained on fully licensed datasets' and that music made with a model relying on an unlicensed dataset 'is not eligible for distribution,' naming Google Flow Music as an approved partner and noting that TuneCore is still building its list of other approved tools.

Rights status

These are two distinct eligibility tests, not one industry standard. DistroKid's stated condition is about the uploader's own rights and conduct; TuneCore's stated condition is about the training-data licensing of the tool itself, a check one level upstream of the uploader. Neither document states that the other's approach is required or standard; this note does not assert that every distributor asks the same question, and a producer working across multiple distributors should expect to answer differently for each.

What to check before you use it

This is an editorial method record describing how to prepare for that question, built from named distributors' own current upload documentation. Before delivery, check each destination distributor's current AI-content policy directly, since the two examples here differ; have the provenance log ready to state which tools were used and, where documented, whether their training data is licensed; and do not assume a release cleared by one distributor's rules would clear another's.

• Does this distributor ask about AI use at all, and if so, about the uploader's rights, the tool's training data, or both?
• Can the session's provenance log answer that specific question without reopening old files?
• If a release is rejected by one distributor's AI policy, would a different distributor's stated rules treat it differently?

A distributor's delivery-time question is only as hard to answer as the records kept during production made it; the two frameworks compared here show that the question itself is not yet standardised across the industry.