A strange division is gradually forming around music.
On one side, we have human-made.
On the other, AI-made.
As if there were a clean line between the two. As if we could attach one small label to a piece of music and instantly understand who created it, how it was made, and how much human work went into it.
The more I work with modern music technology, the less convinced I am that such a line exists.
Perhaps the problem begins with the question itself.
We keep asking:
“Was AI used?”
I think a much better question is:
“What exactly did AI do?”
AI Is Not One Creative Process
Consider a few completely different situations.
A songwriter writes the lyrics, melody and harmony, records the vocals and instruments, and then uses AI-assisted mastering.
Another musician uses AI to restore an old recording.
Someone else uses it to separate stems or remove unwanted noise.
Another artist generates one instrument that becomes part of an otherwise human arrangement.
Someone uses a generator to suggest an initial musical idea and then substantially rewrites it.
And someone enters a prompt and receives an almost complete song.
These are completely different creative processes.
But if all the listener eventually sees is:
AI: YES
what exactly have we told them?
Almost nothing.
Even the Platforms Are Beginning to Recognize That Binary Labels Do Not Work
Interestingly, the industry itself is gradually moving toward the same conclusion.
Spotify has explicitly described AI use in music as a spectrum, not a binary, arguing that the industry needs something more nuanced than forcing every song into “AI” or “not AI.” That is part of the reason it has supported more detailed AI credits through industry standards.
DistroKid now makes another important distinction.
Its AI Credits are intended for cases where AI actually generated part of a track — such as lyrics, composition, vocals or instrumental audio.
Pitch correction, AI-assisted mixing or mastering, and other AI-assisted workflows do not automatically fall into the same category.
YouTube makes a similar distinction between meaningful generation and more limited production assistance such as captions, audio repair, enhancement and certain minor edits.
In other words, platforms themselves are beginning to acknowledge something important:
generated, assisted and processed are not the same thing.
A Simple Example: A Lyric Video
I publish lyric videos.
You can play the song, see the synchronized lyrics and simply sing along with me.
The software may use AI to recognize the words and synchronize them with an existing recording.
Suppose AI performs that task.
What did it create?
The song?
No.
The lyrics?
No.
The melody?
No.
My performance?
No.
It synchronized two things that already existed: the recording and the words.
Calling the result:
“This song was created by AI”
would not make the process more transparent.
It would make the description less accurate.
A much clearer credit would be:
Music & lyrics — human authored. Performance — human. Lyric synchronization — AI-assisted.
Now the listener actually knows what happened.
The Same Problem Exists With Images
Imagine an ordinary photograph of a person.
The person took the photograph.
Software then cleans up the skin slightly, adjusts the lighting or repairs part of the image.
The photograph does not suddenly cease to be a photograph of that person.
This is why some current platform policies already distinguish minor AI-assisted modifications from meaningful generation. YouTube, for example, lists beauty filters and several similar changes among cases that do not require the same disclosure as fully generated content.
That makes sense.
The important question should not be whether AI touched the file.
The important question is what AI actually created or changed.
Now Let’s Make It More Difficult
Imagine another situation.
Someone has never really composed music.
They use a generative system and receive a melody.
They like it.
They transcribe it by ear.
Then they pick up a guitar, play every note themselves, record a real instrument and announce:
“Now it is human-made. I played it myself.”
But performing music and composing music are not the same thing.
A musician can give a magnificent performance of Bach.
That does not make them the composer of Bach.
If the original musical idea came from a generative system, physically replaying it does not automatically relocate the origin of that idea into the performer’s mind.
The musician may genuinely have performed it.
They may have created an excellent arrangement.
They may have transformed the source so extensively that their own creative contribution becomes substantial.
But that is a different conversation.
Authorship cannot be determined by asking:
“Did you play it with your own hands?”
Authorship, Performance and Production Are Different Things
This is where I think much of the current argument begins to fall apart.
We keep mixing three separate questions:
Who conceived it?
Who performed it?
Who produced the final sound?
Those can be three different people.
They always could be.
A composer does not have to be the performer.
A performer is not automatically the composer.
A producer can radically transform the sound of a piece without having written its melody.
AI has simply introduced another participant into an old creative system — a tool that can sometimes perform technical tasks and, in other cases, generate expressive material.
That is why the old categories are no longer enough.
My Problem Is Not With Using AI
I see nothing shameful about using artificial intelligence to develop creativity.
That is one of the things a tool can be for.
If somebody never composed before and a new technology suddenly helps them hear harmonies, experiment with structure, discover ideas and begin creating — why should I consider that a bad thing?
I don’t.
I am genuinely happy when technology opens a creative door that had previously been closed to someone.
Music has always developed alongside tools.
The electric guitar changed our idea of sound.
The synthesizer changed it again.
Sampling caused enormous controversy.
DAWs transformed studio production.
Auto-Tune was once treated almost as the death of “real singing.”
Today, we argue about AI.
That is part of technological change.
But Then Let’s Be Honest About It
This is where I draw a different line.
Not between someone who uses AI and someone who does not.
The more meaningful line is between transparency and pretence.
I have no problem with someone saying:
“Yes, I used a generative tool for inspiration. I struggled to compose before, and now it helps me explore musical ideas.”
Wonderful.
What is much harder for me to understand is someone generating the original musical material, replaying it with live instruments, publicly insisting that they have nothing to do with AI, and then attacking other musicians for using exactly those tools.
At that point, the problem is no longer technology.
The problem is honesty about the creative process.
Copyright Is Beginning to Make the Same Distinction
The legal discussion is also becoming more precise.
The U.S. Copyright Office has stated an important principle: using AI as a tool that assists human creativity does not by itself remove copyright protection from the human work.
Human original expression can still be protected.
Purely AI-generated material is a different matter, and mixed works have to be examined according to the actual human contribution.
Again, the important question becomes not:
“Was AI involved?”
but:
“Where is the human creative contribution?”
Perhaps We Do Not Need One Universal AI Label at All
Instead of one generic warning badge, perhaps we need something much closer to proper credits.
For example:
Human-authored / AI-assisted production The work was created by a person; AI was used as a production tool.
Human composition / AI-assisted performance The composition is human-authored, while certain performances or elements were generated or reconstructed with AI.
Human–AI co-created Generative systems participated in developing musical material while the human directed, selected, altered and contributed original elements.
AI-originated / human-arranged and performed The initial musical material originated from a generative system, while the human created the arrangement and performance.
Predominantly AI-generated Most of the expressive musical material was generated by AI.
That is information.
A single AI badge is barely information at all.
AI Disclosure Should Be Credits, Not a Warning Label
This is probably the simplest way I can express my position.
AI disclosure should function like credits, not like a warning sticker.
We do not put a message at the beginning of a film saying:
“WARNING: COMPUTERS WERE USED IN THE CREATION OF THIS MOVIE.”
We list:
the director;
the cinematographer;
the composer;
the editor;
the VFX team;
the sound department;
the actors.
In other words, we explain who did what.
Why should AI be treated differently?
Not “AI or Human,” but “Where Is the Human and Where Is the AI?”
I think much of today’s conflict exists because we are trying to solve a complex question with an absurdly simple switch.
Human-made.
or
AI-made.
But there is already an enormous space between those two categories.
And that space will only grow.
So I am becoming less interested in asking:
“Did you use AI?”
I want to know:
Who wrote the music?
Who wrote the lyrics?
Where did the original idea come from?
Who arranged it?
Who performed it?
What was generated?
What was processed?
What was reconstructed?
Which creative decisions were made by a human?
That is transparency.
Not shame.
Not an apology.
Not an attempt to prove that one form of creativity is morally superior to another.
Just an accurate description of the process.
Artificial intelligence is a tool.
Sometimes a very small one.
Sometimes an extraordinarily powerful one.
Sometimes it simply cleans up audio.
Sometimes it suggests an idea.
Sometimes it creates a substantial part of a work.
If we genuinely want to understand contemporary creativity, we need to stop throwing all of those situations into a single box marked AI.
Don’t just ask whether artificial intelligence was used.
Ask what it actually did.