Why Your AI-Assisted Song Might Need Two Copyright Registrations

The short answer: When you copyright a song, there are actually two separate things you can protect — the composition you wrote (the lyrics, melody, and arrangement) and the recording you produced (the specific audio file). For most of music history, independent creators didn't need to think about this. But if you're using AI tools in your workflow, the distinction suddenly matters a lot — because the Copyright Office needs to know what's human and what's AI in each one, and those answers are usually different.

This is the part most creators miss. And it's worth understanding before you file.

What's the difference between a PA and an SR registration?

If you've never filed a copyright registration before — or if you have, but it was straightforward — this might be new to you. That's completely normal. The music industry has treated this as inside baseball for decades, and it's only now becoming something independent creators need to think about.

Here's what's happening under the hood:

PA (Performing Arts) protects your musical composition — the underlying song itself. Your lyrics. Your melody. The harmonic structure and arrangement choices you made. Think of it as protecting what you wrote, independent of how it sounds in any particular recording.

SR (Sound Recording) protects the specific recorded version — the actual audio. Your vocal performance, the production decisions, the mix, the way the final file sounds. Think of it as protecting what you captured and produced.

These are legally separate works. They can have different authors, different owners, and — this is the critical part for AI-assisted music — they require different disclosure language when you file with the Copyright Office.

Why does this matter so much for AI-assisted music?

Here's where it gets real. Let's walk through a scenario you'll probably recognize.

You write original lyrics and a melody. You use Suno or Udio to generate instrumental backing tracks. Maybe you record your own vocals on top, or maybe you direct the AI through dozens of iterations until the production matches what you hear in your head. You export the final file. You want to protect your work.

The thing is, the composition and the recording have very different relationships with AI in this workflow:

Your composition probably has a strong human authorship story. You wrote the lyrics. You created the melody. You made the structural decisions about how the song flows. The AI may have influenced the instrumental arrangement, but the core creative choices — the ones the Copyright Office cares about — came from you. When you fill out the Limitation of Claim fields for a PA registration, you're mainly excluding any AI-generated melodic or harmonic material you incorporated without substantially reshaping it.

Your recording is more complicated. The actual audio in that file — the waveforms, the timbral characteristics, the instrumental stems — contains substantial AI-generated content. The Material Excluded field for an SR registration needs to address that directly: the AI-produced instrumental audio, any AI-assisted mastering, and any sonic elements where the tool determined the specific expression rather than you.

See the problem? If you try to cover both the composition and the recording in a single registration with one set of Limitation of Claim language, you're either going to exclude too much (hurting your composition claim) or too little (risking a correspondence letter from the examiner about your recording claim). Neither is a good outcome when you've already paid a non-refundable $65 filing fee.

Can I register both PA and SR for the same song?

Yes — and this is actually what professional songwriters, publishers, and labels have always done, especially when different people are involved in the writing and the recording.

For independent creators who do everything themselves, dual registration has historically felt like overkill. But AI-assisted workflows change that math, and here's why.

The Limitation of Claim requirements are different enough between PA and SR that a single registration may not adequately protect both layers of your work. A composition claim with AI exclusions tailored to the songwriting doesn't address the AI-generated audio content in the recording, and vice versa.

There's also a strategic reason. Your composition (PA) might have very strong human authorship — you wrote every word, every note. Your recording (SR) might have a more complex AI disclosure because of the generated instrumental tracks. Separating the two filings means that a strong composition claim isn't affected by the recording's more nuanced AI story. Each one stands on its own merits.

You don't have to file both. But understanding that you can — and why you might want to — is the kind of thing that protects you from leaving value on the table.

What does dual registration actually cost?

Let's be straightforward about the money, because this is where people understandably hesitate.

Each registration requires a Standard Application with the Copyright Office, which costs $65. For AI-assisted works, the Standard Application is required — not the cheaper $45 Single Application — because you need access to the Limitation of Claim fields where AI-generated content gets disclosed.

So fully protecting a single AI-assisted song with both PA and SR registrations costs $130 in filing fees. For a 10-song catalog, that's $1,300.

Those are real numbers. And they're non-refundable — if a registration gets refused because the AI disclosure was insufficient or unclear, that $65 is gone. Appealing costs another $350 for the first round.

That's exactly why getting the claim language right matters so much. This isn't about perfection for its own sake. It's about protecting a financial commitment you've already made the moment you hit “submit.”

How should the Limitation of Claim language differ between PA and SR?

This is the part that trips up even experienced filers, so let's walk through it carefully. The Copyright Office uses three key fields, and each one should say different things depending on whether you're registering the composition or the recording.

The simplest way to remember it: a PA claim is about what you wrote. An SR claim is about what you recorded and produced. The AI disclosure follows that distinction.

What if I only file one registration?

That's a perfectly valid choice, and many creators will go this route — at least at first. There's no requirement to file both.

Most creators instinctively file SR, because it feels like protecting “the song” — the audio file they created. And it does protect the recording. But here's the tradeoff worth knowing about:

If you only file SR, your lyrics, melody, and arrangement aren't protected as a standalone composition. Someone could take your melodic idea, re-record it without any of the AI-generated instrumental elements, and you wouldn't have a registered composition claim to point to.

If you only file PA, your specific recording — the way the song actually sounds — isn't protected. The composition is covered, but not the production.

For songs where you've invested real creative effort in both the writing and the production, dual registration gives you the most complete protection available under current law. For songs where the composition is more straightforward or the recording is where all the creative work lives, a single registration might be the right call.

The important thing is making that choice intentionally, not by accident.

How does this affect how many Provenance Packs I need?

Because the Limitation of Claim language is different for PA and SR, each registration is its own preparation task — its own Author Created language, Material Excluded language, Note to CO, contributor documentation, and risk assessment. These aren't duplicates of each other. They're genuinely different documents.

A creator filing both PA and SR for a single song would use two Provenance Packs — one tailored to the composition registration, one tailored to the recording. A 10-song catalog with dual registration would need 20 Provenance Packs.

RightsDocket generates the specific USCO form-field language, contributor table, creative decision log, and risk flags for each registration you're preparing. The output is tailored to the registration type so the Limitation of Claim language matches what the Copyright Office expects for that class of work.

You don't need to figure out which fields say what on your own. That's what the Provenance Pack is for.