AI Music Copyright and PRO Registration in 2026

On October 28, 2025, the three largest performing rights organizations in North America did something none of them had done before. ASCAP, BMI, and SOCAN issued a joint announcement that they would each begin accepting registrations for partially AI-generated musical works. For independent songwriters who had been quietly producing with Suno, Udio, and ElevenLabs Music throughout 2024 and 2025, the announcement closed a gap that had defined the industry for two years. AI tracks were finally registrable for performance royalties.

The catch is that the gap did not close on every kind of AI music. There is a specific shape that qualifies, a specific shape that does not, and a paperwork trail you need to keep in order to defend a registration if challenged. This guide walks through exactly what the 2026 PRO rules cover, what you have to document, and the workflow that keeps your AI tracks registrable from day one.

Quick Answer

ASCAP, BMI, and SOCAN now register partially AI-generated musical works as of October 28, 2025, provided the song combines AI-generated elements with meaningful human authorship. Fully AI-generated works (no human creative input beyond a prompt) are not registrable. To qualify, document your creative process (prompts, edits, lyric drafts, structural decisions), disclose AI use during registration, and keep evidence that you guided, edited, or integrated the AI output. Royalty rates are identical to fully human works once accepted.

Key Takeaways

• The October 28, 2025 joint policy from ASCAP, BMI, and SOCAN registers partially AI-generated works.
• Fully AI-generated tracks (with no meaningful human authorship) are still not registrable anywhere.
• You must document the human-led creative process before, during, and after the AI generation.
• Disclosure is recommended at registration and required at distribution under most distributor policies.
• Royalty rates for AI-assisted works are identical to fully human works once accepted.

The October 2025 Policy Change That Opened the Door

For roughly two years before October 2025, the major North American PROs operated in a gray zone on AI music. ASCAP’s registration form did not mention AI at all. BMI quietly declined registrations that mentioned AI tools in the writer credits. SOCAN accepted some AI-assisted works and rejected others on a case-by-case basis with no published guidelines. The result was that thousands of indie producers were registering AI-assisted tracks while saying nothing about the AI, hoping the registration would not be flagged later.

The October 28, 2025 announcement ended that gray zone. ASCAP, BMI, and SOCAN issued aligned policy language confirming that partially AI-generated works are eligible for registration, that registrations will not be treated differently for royalty purposes, and that disclosure of AI use is recommended but not yet mandatory. The press release language was carefully harmonized across all three organizations, a sign that the legal teams had been coordinating for months.

The policy change matters for a few reasons. First, it ended the ambiguity that was keeping cautious songwriters from registering at all. Second, it created a documented standard that distributors could point to when their own AI policies were challenged. Third, it set a precedent that the rest of the global PRO network is now following, with PRS and GEMA already signaling alignment for 2026.

What the announcement did not do is define partial AI generation in legally precise terms. That was left intentionally flexible, which means the burden falls on the songwriter to document what they did versus what the AI did. The next sections walk through how to meet that bar.

What Counts as Partially AI-Generated Under the New Rules

The three PROs all use similar language. A partially AI-generated work is one that combines AI-generated musical content with meaningful human authorship in the final piece. That definition is doing a lot of work and the practical interpretation has been evolving since the announcement.

What clearly qualifies is any song where the human contributes the underlying creative choices and the AI assists with execution. A songwriter who writes the lyrics, sketches the melody, prompts Suno or Udio to generate the production, then edits the output extensively is on solid ground. A producer who writes original lyrics, generates an AI instrumental, rearranges the structure, mixes the result, and records human vocals over the AI bed is even more clearly in the registrable zone.

What does not qualify is a track where the human contribution is limited to writing a single prompt and accepting the AI output as-is. A song produced by typing “upbeat pop song about heartbreak” into Suno, downloading the result, and uploading it under your name is not partially AI-generated under the new rules. It is fully AI-generated, and it remains ineligible for PRO registration.

The middle ground is where most working creators live. A producer who writes a long structured prompt with detailed lyric input, generates several variations, picks the best, then edits the lyrics, swaps a section, masters the result, and ships it. That work is registrable. The key marker is that human creative decisions are visible at each stage of the process. The AI is doing execution. The human is doing direction.

The PROs have not published a percentage threshold. There is no rule that says “thirty percent human” or “sixty percent human” determines eligibility. The standard is qualitative. Does the work demonstrate a human-led creative process? If yes, register. If no, do not.

Human-Led Creative Process: How to Prove It

The single most important thing a 2026 AI musician can do is document the creative process as they work. Not after the fact. As they work. The documentation is what protects your registration if a track gets challenged later, and it also protects you in a copyright dispute downstream.

The documentation that holds up has four components. First, the brief and concept work that preceded any AI generation. This includes notes, lyric drafts, melodic sketches, mood boards, and any pre-existing material you brought to the song. Second, the prompts you fed to the AI tools, with timestamps and version numbers. Third, the editorial decisions you made between AI passes, including the variations you rejected and why. Fourth, the post-generation work, including lyric edits, structural changes, mixing decisions, and mastering choices.

This sounds like a lot of paperwork. In practice it is a single text file per song. I keep a process.md next to every project folder. It starts with my original concept notes (a paragraph), lists every prompt I send to Suno or Udio (with timestamps), notes which variation I selected and why, then logs the edits I make after generation. Five minutes of writing during the session produces a defensible audit trail.

The reason this matters is enforcement. The PROs have signaled that they will not actively audit registrations in 2026, but they reserve the right to do so. More immediately, if a registration is challenged by another rights holder or by a distributor, the documentation is what determines the outcome. Songwriters who can produce a written record of their creative process win those disputes. Songwriters who cannot lose them.

Documentation You Need Before You Register

Beyond the process notes, the documentation the PROs may ask for includes a few specific items. Have these ready before you submit a registration.

The lyric file in its final form, with credit for who wrote what. If you used Claude or ChatGPT to assist with lyrics, the lyric file should be the version you actually sing, after your edits. The prompt history that produced the AI components, exported from your tool of choice (Suno’s history view, ChatGPT’s shared conversation links, Claude’s transcript export). The session log for any DAW work you did, including a brief summary of what you added or changed. The final master, in WAV or high-quality FLAC.

For collaborative songs the documentation gets more involved. Each human collaborator’s contribution should be itemized in writing, with a clear assignment of percentage shares. AI contributions do not get percentage shares, they are not legally a co-author. But the AI-generated material should be explicitly identified, both for transparency and because it affects how the human shares are allocated. The royalties explainer covers the share allocation math in more depth.

Cover art, ISRC, UPC, and metadata can be assembled at distribution time, they are not required for the PRO registration itself. But registering at the same time as you set up distribution is the cleanest workflow, and the distribution checklist covers the full release pipeline. For producers using Melodex to build the full project around the audio (video, social cutdowns, distribution), the documentation flows from the same source files, which keeps the audit trail consistent.

ASCAP, BMI, SOCAN: Where the Forms Differ

The policy alignment between the three PROs is genuine, but the forms themselves are not identical. The practical differences matter when you are filling them out.

ASCAP’s registration form does not yet include a dedicated AI disclosure field, though one is reportedly in development for the 2026 platform refresh. For now, the disclosure is handled in the notes field at the bottom of the form. Use language like “This work was partially generated using Suno V5 with substantial human authorship in lyrics, arrangement, and post-production.” Specific and short.

BMI’s form added an AI disclosure checkbox in November 2025. Check the box, then in the follow-up field describe what AI was used for and what the human contributed. BMI’s review process for AI-assisted works is currently faster than ASCAP’s, with most registrations processing within two weeks versus ASCAP’s three to four.

SOCAN, the Canadian PRO that joined the joint announcement, has the most detailed disclosure form. SOCAN asks for the specific AI tools used, the role of AI in the composition versus the production, and a written statement of human authorship. The form is more work to fill out, but the upside is that SOCAN-registered AI works seem to face the fewest downstream challenges, likely because the documentation is built in from the start.

If you are eligible to register with multiple PROs (you are typically required to pick one for a given work), the practical guidance for 2026 is to default to SOCAN if you are Canadian, BMI if you are American and want faster processing, or ASCAP if you have existing catalog there and want to keep your registrations consolidated.

What the Copyright Office Still Will Not Register

The PRO change is separate from the U.S. Copyright Office position on AI-generated works. The Copyright Office’s policy as of 2026 remains that purely AI-generated works are not eligible for copyright protection, only the human-authored elements of a partially AI-generated work are protectable.

This matters because copyright registration and PRO registration are different things. PRO registration lets you collect performance royalties when your work is played on radio, in venues, on streaming services. Copyright registration lets you sue for infringement and collect statutory damages. They serve different functions and operate under different rules.

The practical implication is that even if your AI-assisted track is registered with ASCAP, you may have limited copyright protection on the AI-generated portions. The lyrics you wrote are yours. The arrangement choices you made are yours. The AI-generated instrumental, melody, and production are not protectable as authored works under current U.S. copyright doctrine.

For most working creators this is acceptable. The PRO royalty stream is what matters for streaming and performance income. The narrower copyright protection on AI-generated portions is mostly relevant if someone copies your track and you want to sue. In that scenario, your case is stronger for the human-authored elements (lyrics, structural changes you made) than for the AI-generated elements.

The U.S. Copyright Office is reviewing this policy through 2026 with potential updates expected in 2027. The Copyright Office’s AI guidance is the source of truth for the latest official position. Until they update, register with the PROs and treat copyright protection as something that covers the human work, not the AI output.

Royalty Math: PRO vs Mechanical vs Sync

Once your AI-assisted track is registered, the royalty math is the same as for any registered work. The PRO collects performance royalties from radio, public venues, streaming services, and other licensees. They take an administrative cut (typically 10-15 percent) and distribute the rest to the registered writers and publishers.

Mechanical royalties are a separate stream. These are paid when a recording is sold, downloaded, or streamed, and they go to the songwriters and publishers (not the recording artists, who get a different stream). In the U.S. mechanicals are administered primarily through the MLC (Mechanical Licensing Collective). MLC accepts registrations for AI-assisted works under similar rules to the PROs, with documentation of human authorship required.

Sync licensing is the third stream and the one with the most variation. Sync royalties come from placements in film, TV, advertising, video games, and other media. Each placement is individually negotiated. Some sync licensors are now AI-friendly, some are not. The major networks and streaming platforms have varying policies, with HBO and FX historically more cautious than Netflix and Amazon. The royalties guide covers the full streams in more depth.

The combined royalty potential for an AI-assisted track in 2026 is real but unevenly distributed. Performance royalties through PROs are paying out normally for registered works. Mechanical royalties through MLC are paying out normally. Sync royalties are case-by-case and slower to materialize. The income mix for a working AI musician today looks much more like a traditional indie musician’s income mix than it did even a year ago.

A Workflow That Keeps Your AI Tracks Registrable

The end-to-end workflow that keeps every track you make registrable is straightforward once you have the habit. Roughly five extra minutes per track. The discipline is to do it every time, not when you remember to.

Start with a written brief before you generate anything. Theme, point of view, target structure, target length. Save it in a file called brief.md in the project folder.

Write or draft your lyrics before you prompt the music tool. Even if you intend to use the music tool’s built-in lyric generation, have your own draft first. The lyric file is one of the strongest pieces of human-authorship documentation you can produce.

When you prompt Suno, Udio, or any AI music tool, save the prompt text and the timestamp. Most tools have a history view, but exporting the prompts to your own file is safer. Note which variation you selected.

After generation, edit. Even minor edits matter. Cut a section, change a transition, swap a chord. Document what you changed in your process.md. The PROs are looking for evidence that you treated the AI output as a draft, not a finished work.

Master your track yourself or with an AI mastering tool, document the choices, and export. Save the original AI output and the mastered version separately, the comparison is good evidence of post-generation work.

Register the work with your chosen PRO, fill out the AI disclosure honestly, and reference the documentation if asked. Set up distribution through DistroKid, TuneCore, or another distributor, fill out their AI disclosure honestly there too. Watch your royalties land. The system works once you work it.

The same documentation discipline applies whether you are building songs alone or producing across the Melodex pipeline. The audio decisions feed forward into the video pipeline, and the documentation forms the audit trail for both.

FAQ

Will I get paid less in royalties for AI-assisted works than for fully human works? No. Once a partially AI-generated work is accepted for registration, the royalty rate is identical to a fully human work. The PROs do not currently differentiate payout based on AI involvement.

Do I need a lawyer to register an AI-assisted track? No. The registration forms are designed to be completed by the songwriter directly. A lawyer is useful if you are managing complex collaborative shares or anticipating a dispute, but for most working solo creators the forms are straightforward.

What if my song was 90 percent AI-generated, can I still register it? There is no published percentage threshold. The standard is qualitative, did the human lead the creative process. A song that is 90 percent AI by audio runtime can still be registered if the human directed the prompts, edited the output, and made the creative decisions. A song that is 50 percent AI by runtime might not qualify if the human contribution was only writing a single prompt.

Are AI vocals treated differently from AI instrumentals? Functionally no, both fall under the same partial AI generation framework. Practically, AI vocal cloning has additional ethics and rights-of-publicity issues if you are cloning someone else’s voice. The voice cloning ethics guide covers the distinction.

Can I register a song that uses an AI-generated sample I licensed from a stock library? Yes. If the sample is licensed for your use, you can build a partially AI-generated work around it. The human authorship in your case is the song you built around the sample, not the sample itself.

What happens if a PRO rejects my AI registration? ASCAP, BMI, and SOCAN all have appeal processes. The most common reason for rejection in 2026 is insufficient documentation of human authorship. Resubmit with a stronger process.md and the rejection often flips.

Does the PRO policy change affect my Spotify or Apple Music status? The PRO policy is separate from streaming platform policies. Spotify and Apple have their own AI disclosure rules, enforced through your distributor. You need to disclose at both layers (PRO registration and distributor metadata) to stay compliant.

Are there countries where AI music is still completely unregistrable? A few PROs in smaller markets have not yet aligned with the ASCAP-BMI-SOCAN policy. JASRAC in Japan has signaled cautious acceptance for 2026. APRA-AMCOS in Australia is reviewing policy through Q2 2026. Most major Western European PROs are aligning. Check your local PRO’s most recent policy statement before assuming acceptance.

What This Changes for Working Creators

The practical takeaway from the 2026 policy change is that AI-assisted music finally has a legitimate path to royalty income. The combination of PRO registration, MLC mechanical registration, and AI-friendly distributors creates a working income stream for indie producers who would have been locked out of formal royalty channels a year ago.

The discipline required is mostly procedural. Document the process. Disclose honestly. Treat the AI as a tool that requires human direction. Register everything. Build the documentation habit into your production workflow so it does not feel like extra work.

Producers who treat the documentation as overhead and try to register undisclosed AI tracks are betting against a tide that is moving against them. The detection tools at the streaming platform layer are improving quickly. The retroactive auditing capability at the PRO layer is being built. The right strategic move in 2026 is to be the producer who has the cleanest documentation, not the one who cuts the most corners. Treat your AI workflow the way you would treat any other professional creative practice.