Learn — By WATMM Desk —

How your AI disclosure travels from your distributor to Spotify and YouTube, and where it gets lost

You never tell Spotify or YouTube how you used AI. You tell your distributor, and it turns your answers into DDEX fields that every store reads in its own way. Here is the path one answer takes, the three places it can lose detail or go missing, and what to check before and after release, as we read the primary pages on 24 Oct 2026.

Key facts

  • The carrier: DDEX ERN 4.3.2 has ContainsAI on each recording, text, image, video and release, and AiContribution on each contributor, both using None, Partly or All.
  • Spotify: AI credits are added through your distributor at upload, are in beta with only certain distributors, and show only on mobile in song credits and Now Playing.
  • YouTube: partners declare Fully, Partly or No Gen AI through DDEX ERN 4.3.2+, a CSV or YouTube Studio; with no value YouTube treats it as unknown and may label the track from its own signals.

If you release music independently, you never tell Spotify or YouTube how you used AI. You tell your distributor, in a form it designed, and the distributor turns your answers into metadata that each store reads in its own way. Spotify says so directly: you add AI credits through your distributor when you upload.

That makes disclosure a relay, and every relay has places where the message changes. This guide follows one answer from your upload form to the listener's screen, names the three places it can lose detail, and gives you a short routine for checking it arrived. If you want the field-by-field detail of the standard itself and a copyable credit template, read our earlier explainer on DDEX ContainsAI and Spotify AI Credits first. This piece is about the journey, not the vocabulary.

Everything below is as we read the primary pages on 24 Oct 2026. Spotify's feature is in beta and the forms change, so check your own distributor before you rely on any detail here.

Stage one: your distributor's form

The first translation happens in a form you didn't design. On DistroKid, the upload flow asks whether any of the track was generated by AI. If you say yes, you pick at least one of four options: the lyrics, the music (AI composed the melody), all of the audio, or part of the audio. Picking "all of the audio" triggers a further question: is your artist name a human or an AI persona?

DistroKid's explainer also draws a line you need to know before you answer. It says pitch correction or auto-tune, AI-assisted mixing or mastering and AI-assisted workflows don't need AI credits; the credits are for generated audio, lyrics and compositions.

The important point for artists is that the form is the narrowest part of the pipe. Whatever detail the standard downstream can carry, the store only receives what the form asked you. Spotify's own newsroom says plainly that not all distributors enable artists to disclose yet. It names a long list of partners working on the standard, including Amuse, Believe, CD Baby, DistroKid, FUGA, RouteNote and SonoSuite. Being on that list tells you a company is involved, not that its upload form already has the questions, so look at the form itself.

Stage two: the DDEX message

Your distributor then sends the release to stores as a DDEX message. DDEX is the industry's shared metadata format, and version 4.3.2 of its release standard (ERN) has two AI fields.

  • ContainsAI sits on each sound recording, text, image, video and on the release. Its allowed values are All ("All content was created using GenerativeAI"), Partly ("Part of the content was created using GenerativeAI") and None ("No content was created using GenerativeAI").
  • AiContribution sits on each contributor. It defaults to All when the contributor is the special GenerativeAI type, and otherwise can be None or Partly.

Notice what this means for your artwork and video. ContainsAI is set per file, so an AI-generated cover image is a separate answer from a human-made recording. If your distributor's form only asks about the track, it is worth asking support how it fills in the image field.

The second loss point is here, and it is our reading of the pages rather than anything the companies say. DistroKid's four options don't map one-to-one onto three values per file plus a value per contributor. "The music" and "the lyrics" describe the composition, while ContainsAI describes files. Someone at the distributor has decided how your ticks become field values, and that mapping isn't published on the pages we read.

Stage three: each store reads it differently

The same message lands at stores that show it in different ways.

Spotify turns it into role credits. Its help page says AI credits are in beta, visible to listeners only on mobile, in song credits and the Now Playing view. They don't appear on artist profiles or as track-level labels. If AI and human contributions exist for the same role, Spotify says to credit them separately. It doesn't explain how to change credits once submitted, and sends you back to your distributor.

YouTube reads the per-file value as a label. Its partner help page says partners declare through DDEX ERN 4.3.2 or later, through a CSV template whose contains_ai column takes all, partly, none or unknown, or in the attribute section of YouTube Studio. Its examples: a prompt-generated download is Fully Gen AI; an AI-generated bassline or string section under your own live vocals and instruments is Partly Gen AI; standard pitch correction on a live vocal is No Gen AI.

This is the third loss point, and the most important one for artists. YouTube says that if you don't provide a value, it treats it as unknown, and it may rely solely on other signals and designate content as fully or partly Gen AI. So on YouTube, silence is not the same as None. A human-made track whose distributor sent nothing is left to YouTube's own judgement. For how that judgement can go wrong, see our piece on how distributors detect AI and the false positives.

Spotify works the other way round. It says that because it depends on artist disclosure, the absence of a credit doesn't mean AI wasn't used. Same missing field, two different meanings. For a wider comparison of the labels, see every platform now labels AI music differently.

A routine that keeps your answer intact

  1. Write your own record first. Before you open the upload form, note each element (lyrics, melody, each vocal and instrument part, mix, master, artwork, video) and whether it was generated, assisted or human. This is the full-resolution version the form will compress.
  2. Read your distributor's definitions. Use its line between generated and assisted, not your own guess. DistroKid's is above; others may differ.
  3. Answer for "no AI" explicitly if the form allows it. Given YouTube's rule on missing values, a clear None is worth more than a blank. Ask your distributor whether it sends None when you answer no, or sends nothing. We could not find this stated on the pages we read.
  4. Ask about artwork and video. If they were generated and the form doesn't ask, tell support and ask how it will be declared.
  5. Check after release. On Spotify, open the track's song credits in the mobile app. On YouTube Music, look at the player page for a label.
  6. Fix it at the source. Corrections go through the distributor, not the store. DistroKid says you can add or update AI credits after upload from the Credits link next to each song on your album page. YouTube's CSV accepts unknown specifically to clear a value that was declared wrongly, which tells you corrections are expected.

This is general guidance, not legal advice.

What we think

The DDEX fields are better than the forms artists actually see. The standard can say "this recording is Partly AI, the strings were generated and the singer is human"; most upload forms let you tick a box or two. Spotify has said the system is not about punishing artists who use AI responsibly or down-ranking tracks for disclosing, and we take that at face value. But as long as the form is coarser than the standard, honest artists will be described less precisely than they described themselves.

The bigger issue is that a blank means different things to different stores. YouTube may fill it in with a detector; Spotify reads it as nothing. Artists who did no AI work are the ones who lose most from that, because they assume there is nothing to declare. We think every distributor should send an explicit None when an artist answers no, publish how its form maps to the DDEX fields, and show artists the AI values it actually delivered for each release. None of that needs a new standard; it needs the middle of the chain to be visible. Until then, keep your own record and check the result after release day.

Sources

  1. ContainsAI (DDEX Data Dictionary, ERN Edition), DDEX (primary source). Accessed 24 Oct 2026.
  2. avs:ContainsAI allowed values (DDEX Data Dictionary, AVS Edition), DDEX (primary source). Accessed 24 Oct 2026.
  3. AiContribution (DDEX Data Dictionary, ERN Edition), DDEX (primary source). Accessed 24 Oct 2026.
  4. AI credits on Spotify, Spotify for Artists (primary source). Accessed 24 Oct 2026.
  5. Spotify Strengthens AI Protections for Artists, Songwriters, and Producers, Spotify (primary source). Accessed 24 Oct 2026.
  6. Gen AI declarations for music partners, YouTube Help (primary source). Accessed 24 Oct 2026.
  7. How to Fill Out AI Credits, DistroKid Help Center (primary source). Accessed 24 Oct 2026.
  8. What Are AI Credits?, DistroKid Help Center (primary source). Accessed 24 Oct 2026.
  9. Spotify to show AI tags in song credits where artists have chosen to disclose through their label or distributor, Music Business Worldwide. Accessed 24 Oct 2026.

Researched with AI assistance, written and checked by WATMM Desk.

Subscribe to We are the music makers

One email a week: the music-tech and AI-in-music news that mattered, and what we think about it.

Sign-ups open soon.