AI Songwriting for the Things You Can't Say Out Loud
People use Donna to write songs for divorces, deaths, apologies they can't make in person, and feelings they can't journal. Here's why AI songwriting works for grief, and how to use it well.

There's a use case for AI songwriting that doesn't get talked about, because it doesn't make for good marketing copy. People use Donna to write songs about things they can't otherwise process.
A divorce. A parent's diagnosis. An apology owed to someone who's stopped taking calls. The thing you almost said at the funeral. The version of the breakup you can't tell your friends because you're embarrassed by your part in it.
These songs almost never get published. They live on someone's phone and get played late at night, alone. And in user surveys, this category — emotional self-expression that the user keeps private — is one of the most-cited reasons people return to Donna week after week.
This post is about why AI songwriting works for things you can't say, and how to do it in a way that actually helps.
Why Songs Work When Other Formats Don't
Journaling is direct. You sit down to write about the thing, and the thing is right there, naked. For some people, in some moments, that's exactly right. For others, it's too much. The defense mechanisms kick in. The page stays blank, or the entry stays surface-level.
Songs work differently. The structure of a song — verses, a chorus, a bridge — gives the brain something to do that isn't "confront this directly." You're choosing words to fit a meter. You're picking a genre. You're naming the song. The technical layer creates just enough distance to let the emotional layer come through.
This is why people who can't journal can sometimes write a song. And it's why people who can journal sometimes find a song shows them something the journal didn't.
The Prompt Mistake People Make
The most common way users approach Donna for an emotional song is also the way that produces the weakest results. They open the app and type something like: "Write a sad song about my divorce."
That prompt will produce a song. It will be technically sad. It will not feel like anything.
The reason: the prompt is abstract. "My divorce" is a category. The brain doesn't store the divorce as a category. It stores it as a Tuesday morning when you found the email. As the dog looking confused. As the box of records by the door. As the specific sentence the lawyer used.
A better prompt names one or two of those specifics. Not the whole story — just an anchor. "Write a song about the morning I realized she'd already moved her things out, including the dog." "Write a song about driving past the house we used to live in." "Write a song about the airport in Phoenix."
The specificity is what produces a song that feels like something instead of a song that sounds like grief.
Write to a Person, Not About a Feeling
A second pattern in the prompts that produce good songs: the user is writing TO someone, not ABOUT a feeling.
"Write a song to my dad telling him I forgive him" produces a different kind of song than "write a song about forgiveness." The first one has a recipient, a stance, a tension. The second one is a Hallmark card.
This is true even when the recipient will never hear the song. Even when the recipient is dead. The brain writes differently when there's a "you."
What to Do With the Song Afterward
Most of the songs people write for things they can't say out loud do not get published. That's fine. That's not a failure of the song.
Some users save them in a private playlist they revisit at anniversaries. Some delete them after listening once and find that the writing was the point — the song was a vessel, and once it carried the thing, it didn't need to exist anymore. Some users send them, eventually, years later, to the person they wrote them for. Some users discover that the song they wrote in private is the most honest piece of work they've ever made and decide to publish it after all.
There's no right answer. The song doesn't owe anyone a destination.
What This Use Case Tells Us
When the team behind Donna started building, the assumed use case was hobbyists making fun songs to share with friends. That use case is real and large. But the use case nobody predicted — people writing songs as a private form of processing — has turned out to be just as common.
It's a quieter use case. It doesn't drive viral TikToks. But it's the one users describe, in surveys, as the reason they keep coming back.
If that's why you're here, you're not alone. And the prompts that work are the specific ones. Name the airport. Name the morning. Write to the person, not about the feeling. The song will know what to do.
Ready to create your own song?
Turn any idea into a full track with Donna's AI music generator — in minutes, no experience needed.
Try Donna Now