# AI disclosure when you release music: what you actually have to declare > On Spotify's current beta, participating distributors let you disclose AI-generated contributions by role rather than label the entire track. Lyrics, vocals, instrumental performances, and production are common examples, and human and AI contributions can be credited separately for the same role. Availability and wording still depend on your distributor, so its live release form is the source of truth when you upload. _By [Chronatum](https://chronatum.com/) · Published 2026-08-06, updated 2026-08-08_ Most confusion about AI disclosure comes from imagining it as a verdict on the whole track. It is not. It is a set of component-level statements you make about how the recording was produced, and a mixed answer — AI on one element, none on the others — is the case the system was designed for. ## Where the declaration actually happens For Spotify's beta, you declare through a participating distributor when you upload, and that information can appear in Song Credits and the Now Playing view on mobile. Spotify is also working with DDEX on a broader industry standard so role-level disclosures can travel consistently through the music supply chain. The feature is still evolving and is not available through every distributor. That has a practical consequence worth internalising: **the declaration is metadata you supply, not an inference drawn from your audio.** Detection systems exist and run separately (that is what produces [an AI flag on a track you made yourself](/learn/song-flagged-as-ai)), but the disclosure record itself is what you said, attached to the release. > **Note.** Policies here are moving quickly and this page is a starting point, not a compliance document. As of August 2026 Spotify describes AI credits as an optional beta available through certain distributors. Check your distributor's current help documentation before you release — its form is the authority on what you are submitting. ## Role-level credits, not a whole-track label Spotify currently gives these as common examples of roles that can carry an AI credit: - **Lyrics** — words generated or co-written with a model. - **Vocals** — a generated vocal or AI-created voice performance. - **Instrumental performances** — generated instrumental parts or backing. - **Production** — an AI-generated contribution credited in the production role. Because credits are role-level, a track can carry both human and AI contributions without being reduced to an "AI song" or "not AI" label. Do not extrapolate beyond the fields your distributor actually shows: a tool or service may use different categories while the beta develops. ## The grey areas, and a rule for them The genuinely hard question is where ordinary studio tooling ends and a declarable AI component begins. Nobody expects you to declare a compressor. But the line has moved, and several things sitting in most sessions now are somewhere near it: stem separation, adaptive mastering services, generative sample and loop tools, voice conversion, and the newer assistants built into DAWs. A workable rule: **declare when a tool produced creative content, not when it processed content you produced.** A limiter shaping your mix is processing. A model generating a bassline, writing a couplet, or synthesising a vocal take is producing. Pitch correction on your own performance is processing; converting that performance into a different singer's voice is producing. Where you cannot decide, declare. The asymmetry strongly favours it — see below. ## Why over-declaring is cheaper than under-declaring An accurate declaration of AI use is a metadata field. It does not currently remove a track from streaming services, and platforms have generally framed these tags as context for listeners rather than as a penalty. Whether a particular audience cares is a real question, but it is a marketing question. An inaccurate declaration is a different category of problem. It is a false statement attached to a release, made in a form you accepted terms on, and it is the kind of thing that follows a catalogue rather than a single track. If a dispute later turns on your credibility — and disputes about AI usually do — one bad declaration is worth more against you than any detector score. So the practical posture is: declare accurately, declare consistently across releases, and when a component is ambiguous, resolve the ambiguity toward disclosure. ## What disclosure does not do Declaring no AI on any component is a statement by you. It is not verified by anyone at the point you make it, and it does not by itself stop a detector from flagging the recording later — the two systems are independent, which is why artists who used no AI at all still occasionally have to answer for a flag. That gap is the reason to keep your own record of how a track was made. A declaration says what you assert; a contemporaneous record of the session is what you can show when somebody asks you to support it. [What counts as evidence, and what does not](/learn/how-to-prove-you-made-your-music) goes into that in detail. ## Sources and further reading - [AI credits on Spotify](https://support.spotify.com/us/artists/article/ai-credits/) — Spotify for Artists - [Spotify Strengthens AI Protections for Artists, Songwriters, and Producers](https://newsroom.spotify.com/2025-09-25/spotify-strengthens-ai-protections/) — Spotify - [DDEX standards](https://ddex.net/standards/) — DDEX --- More: [llms.txt](https://chronatum.com/llms.txt) · [sitemap](https://chronatum.com/sitemap.md)