All articles

Disclosing AI Use During Distribution: DistroKid's Upload Step, Spotify's AI Credits and the Declaration After a Re-Record

The first time DistroKid asked me whether AI generated any part of the track I was uploading, I sat with my cursor hovering over the answer far longer than I expected to. Not because the question was hard, but because I suddenly understood that this small form was going to travel with the song onto Spotify, into the public credits, in front of every listener who taps the song info panel.

Here is the short version before we go deep. If you distribute through DistroKid, you will now hit an AI Credits step during upload that asks whether AI generated the lyrics, the vocals, the instrumental performance, or the whole composition. Whatever you tick gets packaged into your release metadata and delivered to Spotify, which started showing AI disclosures inside Song Credits on mobile in a beta that launched on April 16, 2026 with DistroKid as the first partner. Disclosure is voluntary, it does not hurt your recommendations, and you can edit it after release. And if you re-record an AI-generated song with real humans, the declaration does not simply disappear, which is the part almost nobody explains properly. I will.

I have been disclosing AI use during distribution since the option existed, across tracks that range from fully human to heavily AI-assisted, so what follows is a practitioner’s walkthrough rather than a press release summary. Every number and date here comes from research done for this article, and I will name sources as we go.

Why Everyone Suddenly Wants a Declaration

None of this appeared out of thin air. In September 2025, Spotify announced a three-part cleanup covering impersonation, spam filtering, and standardized AI disclosures, and revealed in the same breath that it had removed more than 75 million spammy tracks in the previous 12 months. Music Business Worldwide, which covered that briefing, also noted Deezer was receiving over 30,000 fully AI-generated tracks per day at the time. That is the flood the industry is reacting to.

The important nuance, and the one that should lower your blood pressure, is that the crackdown targets fraud, not honest AI use. As TechCrunch reported when Spotify updated its AI policy, the disclosure system was framed as transparency for listeners, not a punishment mechanism for creators. Dynamoi’s analysis of the policy makes the same point: the spam filter goes after mass uploads, duplicate dumps, metadata manipulation, and artificially short tracks, none of which describes a person carefully releasing music they made with AI assistance.

So the declaration exists because platforms need a way to separate the two groups. People who disclose look like the first group. People who hide and get caught look like the second. That framing should drive every choice you make in the upload form.

What the DistroKid Upload Step Actually Asks

DistroKid rolled its AI disclosure questions into the upload flow itself, and Digital Music News, which documented what the DistroKid AI credit disclosure looks like, described the broad launch in mid-May 2026 as still expanding, with the questions appearing only on new uploads rather than being applied retroactively across your back catalog.

The step is simple on the surface. You are asked whether any part of the track was generated by AI. If you answer yes, you specify which elements. DistroKid’s own help documentation lists the categories in plain language: the lyrics, meaning AI wrote the words; the music, meaning AI composed the melody; and further options covering AI-generated vocals and instrumental performance, up to a fully AI-generated track. This mirrors the structure of the industry standard underneath, which I will get to in a moment, because those categories are not DistroKid’s invention.

Two things caught me off guard the first time through. First, the form treats these as separate declarations, not one big AI stamp. A track with human vocals over an AI-generated instrumental gets a different declaration than a track where AI wrote the lyrics but humans played everything. That granularity is the whole point, and it is worth slowing down to answer accurately rather than ticking the first box that seems close enough.

Second, the system is entirely self-reported. Music In Africa’s coverage of the rollout stressed that DistroKid relies on voluntary disclosure rather than automated detection, which also means the absence of an AI credit on someone else’s track proves nothing. Deezer went the other way and built automated tagging. DistroKid and Spotify chose to trust you, at least for now, and I would not bet on that trust remaining unverified forever.

The Line Between an AI Tool and AI Generation

This is where most creators get stuck, so let me give you the working rule I use. The declaration is about generation, not assistance. If AI produced creative content that survives into the release, you declare it. If AI helped you polish content that humans created, you generally do not.

DistroKid’s guidance draws this line explicitly. As Jack Righteous notes in his guide to Spotify AI Credits and DistroKid uploads, DistroKid’s documentation says you do not necessarily need AI Credits for pitch correction, auto-tune, AI-assisted mixing, or AI-assisted mastering. Those are tools operating on human material. Using a mastering service on your self-recorded song does not make the song AI-generated any more than using a spell checker makes an essay AI-written.

Flip it around and the rule holds. Suno wrote your instrumental and it is in the final mix, that is AI-generated instrumentation, declare it. ChatGPT wrote your lyrics and you sing them, the audio is human but the words are not, declare the lyrics. You hummed a melody, played every note yourself, and only used AI to clean up the low end, declare nothing.

The gray zone I get asked about most is AI as a writing partner. If you generated twenty lyric drafts, kept two lines, and rewrote everything else yourself, I would still tick the lyrics box if those two lines are recognizably in the song, then relax knowing the credit does not carry a penalty. Honesty is cheap right now. Getting caught misrepresenting later, when detection improves, will not be.

How Your Answer Becomes a Spotify AI Credit

Once you submit, your declaration does not sit in a DistroKid database somewhere. It gets encoded using a metadata standard from DDEX, the Digital Data Exchange, which is the not-for-profit standards body that already defines how distributors deliver track data like ISRC codes and songwriter splits to streaming services. Billboard’s reporting on the Spotify launch confirmed the AI credits system is being developed through DDEX with the stated goal of industry-wide alignment rather than a Spotify-only label.

Dynamoi’s technical breakdown of how DDEX AI disclosure works on Spotify describes five disclosure categories in the standard: AI-generated vocals, AI instrumentation, AI composition, AI post-production, and AI lyrical content. Notice how closely DistroKid’s upload questions map onto those. That is deliberate. Your single declaration at upload is designed to travel through the whole supply chain, which means the same disclosure could eventually surface on Apple Music, Amazon Music, or YouTube Music without you filling out anything new. Apple has already begun phasing in its own Transparency Tags for label deliveries as of March 2026, according to RightsDocket’s review of the disclosure landscape, so this is clearly the direction of travel.

The practical takeaway is that you are not answering a DistroKid quiz. You are creating a permanent, structured, machine-readable statement about your song that platforms you have never thought about may display years from now. Treat it with the same care you give your songwriter credits.

What Listeners Actually See

On Spotify, the result of all this plumbing is modest, and honestly that surprised people who expected a scarlet letter. During the beta, the AI disclosure appears inside the Song Credits panel on the Spotify mobile app. Chartlex, which analyzed the feature from the indie artist side in its review of Spotify’s AI song credits disclosure, reports the tag is only visible on mobile during the beta, there is no desktop surface yet, there is no separate AI music section, and there is no warning label or demotion attached.

Spotify has been unusually direct about the algorithm question because it knows that is the fear stopping people from disclosing. The company has stated that AI disclosures do not affect how music is treated for recommendations, a point repeated by Spotify staff in its own community forums when listeners asked for AI filtering. Chartlex also passes along Spotify’s caveat from the other direction: the absence of a tag does not mean a track is AI-free, since the whole thing is voluntary.

My own experience matches the official line. Tracks where I disclosed AI instrumentation have performed no differently in algorithmic playlisting than comparable undisclosed human tracks from the same profile. The sample size of one artist proves nothing scientifically, but I have also not found a single credible report of disclosure tanking a release, and believe me, if it were happening, the forums would be on fire.

The Re-Record Question Nobody Answers Clearly

Now the part of the title that pulled you in, and the scenario I have lived through personally. You generate a song with AI, love it, and then re-record it from scratch. Real drummer, real guitars, your own voice, human engineer. Not one second of AI audio survives into the new master. Do you still declare AI use?

Yes, partially, and here is the logic. A song is two things stacked on top of each other: the recording and the underlying composition. Your re-record replaced the recording. It did not replace the composition. If AI composed the melody or wrote the lyrics, those elements are still AI-generated no matter who performs them, the same way a cover of a Beatles song does not make you the songwriter. This is exactly why both DistroKid’s form and the DDEX standard split composition and lyrics into their own categories, separate from vocals and instrumental performance.

So for a full human re-record of an AI-written song, my declaration looks like this: AI-generated composition yes, AI-generated lyrics yes if applicable, AI-generated vocals no, AI-generated instrumentation no. The credit that surfaces tells the truth: humans performed this, a machine wrote it. Several guides get this wrong by treating disclosure as purely an audio question. The audio test matters for the performance categories, but composition and lyrics survive any number of re-records.

There is also the reverse timeline, which trips people up in a different way. Say you released the AI demo version first, disclosed it fully, and later replace the audio with your human re-record, something DistroKid supports as an audio replacement feature on its top plan. The moment the master changes, your old declaration is wrong. The vocals and instrumentation credits now overstate the AI involvement, and leaving them in place undersells the very human work you just paid for. Update the declaration to match the new master, keeping only the composition and lyrics credits that still apply.

One warning while we are here, because I have seen this pitched in forums as a clever move. Re-recording purely to escape the AI label, then declaring nothing at all, is misrepresentation if the melody and words came from a machine. It also throws away the honest middle position that actually helps you, which is telling listeners that real musicians brought an AI sketch to life. That is a good story. Hiding it turns a good story into a liability.

Editing a Declaration After Release

The single most useful thing I can tell you about the whole system is that the declaration is not carved in stone. DistroKid’s help center spells out how to fill out AI credits and confirms you can add or update AI credits after uploading by going to your album page, clicking Credits next to each song, or visiting the credits page directly at distrokid.com/credits.

That matters in three real situations. First, the honest mistake: you rushed the upload, answered from memory, and realized later that one track on the EP had an AI-generated bridge you forgot about. Fix it in the credits editor and the corrected metadata flows out. Second, the back catalog: DistroKid’s rollout only questions new uploads, but if you have older releases with AI involvement, you can disclose retroactively through the same credits page rather than waiting to be asked. Chartlex confirms older releases can be updated through a metadata refresh and that DistroKid lets you edit AI status on existing tracks. Third, the re-record replacement I just described, where the composition of the audio changed and the declaration needs to follow it.

My habit now is to review credits about a week after every release goes live, once the upload adrenaline has worn off. It takes five minutes and it has caught two errors across my catalog so far. Cheap insurance.

Honest Problems With Self-Reporting

I like this system and I still want to be straight with you about its weaknesses, because trusting it blindly will burn you in a different way than ignoring it would.

The obvious one is that self-reporting only captures the honest. Digital Music News called this out on launch day: bad actors are exactly the people least inclined to disclose, so the tags currently tell you more about a creator’s integrity than about the actual prevalence of AI music on Spotify. That asymmetry will not last. Deezer already runs automated AI detection and tags tracks without asking anyone, and both DistroKid and Spotify have left the door open to changing their approach. When automated verification arrives, the historical record of who disclosed honestly and who did not will already exist, written by everyone’s own hands.

The second weakness is inconsistency across distributors. Disclosure norms differ depending on who carries your music, with some distributors collecting more granular attribution and at least one major distributor known for rejecting fully AI-generated content outright. If you distribute the same catalog through multiple pipelines, your disclosures can end up uneven through no dishonesty of your own. Keep your own record of what you declared and where.

The third is the beta itself. Mobile-only visibility, one distributor at launch, and an industry standard still being finalized through DDEX means details will shift under your feet. RightsDocket, which reviews this space against contemporaneous reporting, flags the formal DDEX AI disclosure standard publication as the next thing to re-check. I would treat any specific screenshot of the flow as provisional and the underlying principle, disclose generation honestly by category, as stable.

And one blunt warning. There are paid services openly marketing the removal of AI detection fingerprints from generated audio so tracks can pass platform screening undisclosed. Stay away from them. Whatever the marketing says, the product is deniability, and buying deniability about your own music is about the strongest signal of bad faith a platform could ever ask for if things go wrong later.

What It Costs to Do This Through DistroKid

Disclosing AI use costs nothing extra. The AI Credits step is part of the normal upload flow on every plan. What you pay for is the distribution itself, and it is worth knowing the real numbers because DistroKid’s headline price is not the whole story.

According to Chartlex’s breakdown of DistroKid pricing in 2026, the three plans are Musician at 24.99 dollars per year for one artist name, Musician Plus at 44.99 dollars per year for two artist names plus custom release dates and daily stats, and Ultimate at 89.99 dollars per year for up to 100 artist names. All plans include unlimited uploads with no commission on streaming royalties. The add-ons are where budgets quietly inflate: Leave a Legacy is a one-time 29 dollars per single or 49 dollars per album to keep music live if you ever stop subscribing, and YouTube Content ID runs 4.95 dollars per single per year with a 20 percent revenue share, which for AI-heavy catalogs is often money badly spent anyway given how contested AI content is inside Content ID.

Two plan notes specific to AI workflows. The base Musician plan does not include custom release dates, which matters if you are coordinating a disclosure-updated re-release. And the audio replacement feature relevant to the re-record scenario sits on the Ultimate plan, so factor that in if swapping masters is part of your plan rather than a one-off.

Three Questions That Settle Any Declaration

When a track sits somewhere in the messy middle and you genuinely do not know what to tick, I run it through three questions in order. They resolve every edge case I have hit so far, including collaborations where different people used different tools.

Did AI produce any audio that a listener can hear in the final master? If yes, the performance categories are in play. Trace each audible element back to its origin: the vocal take, the drums, the pads, the topline. Anything a model rendered gets its category ticked, even if you chopped it, pitched it, and buried it under three other layers. Processing AI audio does not turn it into human audio.

Did AI write words or music that humans then performed? This is the re-record question generalized, and it catches the cases the audio test misses. A session singer delivering lyrics that came out of a chat window, a string quartet reading a chart that a model composed, both are human recordings of AI writing. Composition and lyrics get declared, performance categories stay clean, and the credit ends up telling exactly that story.

Would I be comfortable if my declaration were printed next to an interview where I described how the song was made? This one is less technical and more of a gut check, but it is the question that actually protects you. The most avoidable damage in this whole area is not a platform penalty, it is the screenshot of an artist denying AI use next to evidence of it. Your DistroKid declaration, your social posts, and your press answers should all describe the same song. If they do, there is no story. If they do not, you wrote the story yourself.

Run those three in order and the form fills itself out. The honest answer is almost never the scary one; the inconsistent answer is.

The Workflow I Actually Use

Everything above condenses into a routine that takes maybe fifteen minutes per release and has never once created a problem for me.

Before opening DistroKid, I write a one-page release note per track: which tool generated what, which elements survived into the final master, who the human contributors are, and what I will therefore declare. I settle the declaration outside the upload form, in writing, while I can still think. Deciding inside the form from memory is how people end up with credits that contradict their own social posts about how the song was made.

During upload, I answer the AI Credits step from that note, category by category. Lyrics, composition, vocals, instrumentation, each one gets a deliberate yes or no. If a category is genuinely ambiguous, I disclose it, because the disclosure carries no algorithmic cost and the misrepresentation carries plenty of future risk.

After release, I check the Spotify mobile app about a week later, open the song credits, and confirm the AI credit renders the way I intended. Then I file the release note with a screenshot. If I ever re-record or replace audio, the note gets updated first, then the credits at distrokid.com/credits, in that order, so my paper trail and my public declaration never drift apart.

That last habit, the paper trail, is the one I would push hardest. Disclosure tells platforms that AI was involved. It does not document what the humans did, and if a rights question ever lands on your desk, the human contribution record is what you will wish you had kept.

What I Am Watching for the Rest of 2026

Three developments will decide how this system matures, and I would keep half an eye on each before making long-term catalog decisions.

More distributors joining the Spotify beta is the nearest one. DistroKid was the launch partner, other pipelines are expected to follow, and once the majors and remaining indie distributors light up, disclosed AI credits stop being a DistroKid quirk and become the default state of new releases. The share of disclosed releases is already climbing quarter over quarter according to Chartlex’s campaign data, even though it remains a small fraction today.

The formal DDEX standard publication is the second. Once the specification is finalized, expect the categories to firm up and expect Apple, which began phasing in Transparency Tags for label deliveries in March 2026, to converge on the same metadata rather than maintaining a parallel system forever.

Regulation is the third and the one with teeth. Silverman Sound’s guide to AI music legal risks notes that EU rules on marking AI-generated audio become enforceable in August 2026, which moves disclosure from platform etiquette toward legal obligation for anyone distributing into European markets. Voluntary systems built now are, in my read, the industry getting its paperwork in order before the law asks to see it.

Put all of that together and the conclusion writes itself for anyone releasing music this year. Disclosing AI use during distribution is currently free, penalty-free, editable, and structured in your favor. Every trend line points toward it becoming verified, standardized, and mandatory. The creators who will be fine in 2027 are the ones whose declarations were already accurate in 2026. This week, open distrokid.com/credits, walk your existing catalog, and make every AI credit match reality, including the re-records. It is the least glamorous hour you will spend on your music this year, and possibly the most protective.

Sources

Common questions

What does DistroKid's AI Credits step ask?

Whether any part of the track was generated by AI, and if so which elements: the lyrics, the music or melody, AI generated vocals, AI instrumental performance, or a fully AI generated track. Each is a separate declaration rather than one AI stamp, so a track with human vocals over an AI instrumental gets a different answer than one where AI wrote the words but humans played everything. The questions appear on new uploads only and are entirely self reported.

Do I need to declare AI use for mastering or pitch correction?

No. DistroKid's guidance says AI Credits are not needed for pitch correction, auto tune, AI assisted mixing or AI assisted mastering, since those are tools operating on human material. The declaration is about generation, not assistance. If AI produced creative content that survives into the release, declare it. If AI helped polish content humans created, you generally do not.

Does disclosing AI use hurt my Spotify recommendations?

No. Spotify has stated that AI disclosures do not affect how music is treated for recommendations, and Spotify staff have repeated this in community forums. There is no warning label, no demotion and no separate AI section. The author's disclosed tracks have performed no differently in algorithmic playlisting than comparable undisclosed human tracks from the same profile.

Where do listeners see the AI credit on Spotify?

Inside the Song Credits panel on the Spotify mobile app, which is the only surface during the beta. There is no desktop display yet. Because the system is voluntary, Spotify also notes that the absence of a tag does not mean a track is AI free.

If I re record an AI song with real musicians, do I still declare AI?

Yes, partially. A song is a recording stacked on a composition, and your re record replaced only the recording. If AI wrote the melody or lyrics, those elements remain AI generated no matter who performs them, just as covering a song does not make you its writer. The correct declaration is AI composition yes, AI lyrics yes if applicable, AI vocals no, AI instrumentation no. Re recording purely to escape the label and declaring nothing is misrepresentation.

Can I change my AI declaration after release?

Yes. DistroKid's help center confirms you can add or update AI credits after uploading by going to your album page and clicking Credits next to each song, or by visiting distrokid.com/credits directly. This covers honest mistakes, retroactive disclosure on older releases that were never asked, and audio replacements where a human re record makes the old vocal and instrumentation credits overstate AI involvement.

How does my DistroKid answer reach Spotify and other platforms?

It is encoded using a DDEX metadata standard, the same body that defines how ISRC codes and songwriter splits are delivered. The standard has five categories: AI vocals, AI instrumentation, AI composition, AI post production and AI lyrical content, which DistroKid's questions map onto. The same disclosure is designed to travel the whole supply chain, and Apple Music began phasing in its own Transparency Tags for label deliveries in March 2026.

What are the weaknesses of self reported AI disclosure?

It only captures the honest, so current tags say more about a creator's integrity than about how much AI music exists. Norms differ between distributors, with at least one major distributor rejecting fully AI generated content outright, so disclosures can end up uneven across pipelines. And the beta is still shifting, with mobile only visibility and the DDEX standard not yet formally published. Deezer already runs automated detection, and verification is expected to arrive elsewhere.

How much does DistroKid cost?

The AI Credits step itself is free on every plan. Distribution runs $24.99 a year for Musician with one artist name, $44.99 for Musician Plus with two names, custom release dates and daily stats, and $89.99 for Ultimate with up to 100 names. Add ons include Leave a Legacy at $29 per single or $49 per album, and YouTube Content ID at $4.95 per single per year with a 20 percent revenue share. Audio replacement, relevant to re records, sits on the Ultimate plan.

How do I decide what to declare on a borderline track?

Ask three questions in order. Did AI produce any audio a listener can hear in the final master, which triggers the performance categories even if you chopped or buried it. Did AI write words or music that humans then performed, which triggers composition and lyrics. And would you be comfortable if your declaration were printed next to an interview about how the song was made. The honest answer is rarely the scary one, the inconsistent answer is.