Do I legally have to disclose AI-generated music?

There is no single global rule. Obligations come from the platform you publish on and, increasingly, from regional transparency law, not from a universal music standard. In practice the disclosure that matters is the answer you give your distributor at upload, because that is the field that travels into store metadata.

Will disclosing AI involvement hurt my release?

It can affect discovery on services that classify AI uploads, and it does not usually cause removal. An inaccurate answer is the more expensive path: metadata disputes and account-level action are far worse outcomes than a track that gets less editorial support.

What counts as generative AI in this context?

Tools that produce new musical material or a synthetic voice. Pitch correction, stem separation, noise reduction, amp modelling and mastering assistants are processing, not generation, and every platform treats them as ordinary production.

Does this page store what I type?

No. The statement is assembled in your browser as you type. Nothing is uploaded, logged or retained, exactly like the detector itself.

Can a detector prove my disclosure is honest?

No, and that is the point of disclosing. Acoustic analysis gives probabilistic evidence about a recording, never a provenance record. Accurate metadata and retained project files are the only durable proof of how a track was made.

Free tool

AI music disclosure generator

Answer three things about how the track was made and get wording you can paste into a distributor field, a credits note, a YouTube description or a release page. Built in your browser; nothing you type leaves the device.

Short answer

How do you disclose AI-generated music?

Disclose at the point the question is asked: the AI-involvement field in your distributor's upload flow, accurate credits for every human contributor, the altered-or-synthetic toggle on YouTube where it applies, and a plain sentence in the release description. Say what the model produced and what a person did, and keep your project files.

How each platform asks →

1. What did the AI do?

No generative AIStandard production tools only — tuning, stem separation, mastering assistants, amp models.AI-assisted writing or productionA human performance and arrangement, with generative help on ideas, sounds or sections.AI or cloned vocals over human musicThe instrumental was produced by people; the voice is synthesised or converted.AI instrumental with human vocalsThe backing track came from a generator; the vocal is a real recorded performance.Fully AI-generated trackMusic and vocals both produced by a generator, with prompt and curation work by you.

2. Tool or model used (optional)

3. What did you do? (optional)

4. Where is it going?

Your disclosure

This recording is AI-assisted: generative tools contributed to parts of the writing and production, and the arrangement, performance and final mix decisions are human. All rights required for this release are held or licensed, and project files are retained in case the release is queried.

Paste this into the AI-involvement or notes field, and keep the credit list itself accurate for every person who contributed. Your distributor is where disclosure actually enters industry metadata.

Why the wording matters more than the label

Almost every platform dispute we see described publicly is not about whether a generator was involved. It is about a mismatch: a release whose metadata says one thing and whose audio, credits or artist name suggest another. Saying plainly what the model produced and what a person did removes the mismatch, and it costs a single sentence.

Note what a disclosure is not. It is not a detector result, and a detector result is not a substitute for it. Our own analysis returns a probability from measurements of the audio, which is evidence, not provenance — see what a result actually means and where the limits are. The only durable record of how a track was made is the one you keep: session files, prompt history, stems, dated exports.

Where each platform asks

The distributor upload flow is the field that matters, because that answer travels into store metadata. YouTube asks separately in the upload flow and has its own route for singing-voice likeness. Some services classify uploads themselves and adjust discovery without asking you at all. The current shape of all of that is on AI music rules by platform.

Questions about disclosure

Do I legally have to disclose AI-generated music?: There is no single global rule. Obligations come from the platform you publish on and, increasingly, from regional transparency law, not from a universal music standard. In practice the disclosure that matters is the answer you give your distributor at upload, because that is the field that travels into store metadata.
Will disclosing AI involvement hurt my release?: It can affect discovery on services that classify AI uploads, and it does not usually cause removal. An inaccurate answer is the more expensive path: metadata disputes and account-level action are far worse outcomes than a track that gets less editorial support.
What counts as generative AI in this context?: Tools that produce new musical material or a synthetic voice. Pitch correction, stem separation, noise reduction, amp modelling and mastering assistants are processing, not generation, and every platform treats them as ordinary production.
Does this page store what I type?: No. The statement is assembled in your browser as you type. Nothing is uploaded, logged or retained, exactly like the detector itself.
Can a detector prove my disclosure is honest?: No, and that is the point of disclosing. Acoustic analysis gives probabilistic evidence about a recording, never a provenance record. Accurate metadata and retained project files are the only durable proof of how a track was made.

This page is orientation, not legal advice. Platform terms change; read the current version before a release you care about.