The Robot in the Studio Would Rather You Didn't Ask

In March, Rolling Stone's Nathan Brackett published "Inside the Don't Ask, Don't Tell Era of AI in Music," and the title came straight from one of his sources. "People don't really admit to what extent they're using it," songwriter Michelle Lewis told him. She describes the mood among her peers as "Don't ask, don't tell."

We think it's the most useful thing written on the subject this year, mostly because it isn't a hot take. It's plain reporting from the people who actually make commercial music.

What the pros are actually doing

The article cites a survey by the audio-tech company Sonarworks of more than 1,100 producers, engineers, and songwriters. Seven out of 10 said they were at least occasionally experimenting with AI tools. One in five were regular users.

That's a survey, not an audit of the whole industry. But it does tell you the tools have moved well past the curious-hobbyist stage.

Most of that use is less dramatic than the headlines suggest. According to the piece, people mostly use AI for narrow, time-saving jobs: restoring audio, pulling vocals and instruments apart (called stem separation), and mastering. Lumineers producer David Baron calls stem separation "phenomenal." Songwriters in Nashville and Los Angeles use tools like Suno to turn lyrics and chords into polished demos. "In private, songwriters are saying, 'It's kind of awesome,'" Lewis says.

Some of it goes further. Jay-Z's longtime producer and engineer Young Guru guesses that "more than half" of sample-based hip-hop now uses AI-made funk and soul samples instead of licensed records or hired players. That's his estimate, not a measured figure, but coming from him it's hard to wave off.

Then there's Charlie Puth, who used a tool called Replay to test ideas for his upcoming album. "There's a setting [that makes] a mono vocal sound like eight to 10 people singing," he told Rolling Stone. "I'm gonna use that to see if I even want the choir sound on there. And then we take it off and replace it with a real choir." He calls that "the correct way of using AI." Asking a machine to remake a finished song from scratch, he says, is "nauseating."

We like that example because it shows what most of this looks like: AI as a sketchpad, with humans doing the final performance.

When the sketchpad ends up on the charts

Sometimes the sketch becomes the record.

In July, the ABC reported that Australian DJ and producer Josh Fawaz had added AI credits on Spotify to his cover of Madonna's "Like a Prayer." He added them after an ABC investigation, by which point the song had already held the No. 1 spot on Australia's national airplay chart for four weeks and had been streamed more than 38 million times on Spotify. The new credits list generative AI vocals and AI drums.

Then came the summer's strangest hit. Fenix Flexin, best known for trap music with Shoreline Mafia, released "Rubberz," an '80s-style synth-pop song sung in a voice that, as WIRED put it, sounds "closer to a very chipper Morrissey." It climbed to No. 58 on the Billboard Hot 100.

Fans suspected AI. Fenix denied it, saying the only differences were "auto-tune reverb and me using my fake uk accent." Then producer Medasin posted a video tying the song to a little-known AI music app called Treblo. Treblo's own new detector rated "Rubberz" "Very likely Treblo." That detector hasn't been independently reviewed, so treat it as strong evidence, not a verdict.

According to the Guardian, Fenix eventually admitted using AI, writing that he "never said I didn't use AI, i said it had nun to do w recording process." Exactly what the machine did and what the human did is still disputed. That fuzziness is sort of the whole problem.

Two revolutions, running at different speeds

As we see it, the music business is living through two revolutions at once.

The first is technical: how music gets made. That one is already well underway, and the survey Rolling Stone cites makes that hard to argue with.

The second is cultural: whether anyone will admit it. That one is mostly a public-relations fight. Baron told Rolling Stone there can be a "social penalty" for using AI. When the penalty for telling the truth is higher than the penalty for staying quiet, people stay quiet. We don't think that's a mystery.

The industry is starting to push back on the silence. In July, IFPI, the RIAA, the Grammys, SAG-AFTRA and others announced voluntary track labels separating "AI-Generated" recordings from "AI-Assisted" ones. Under the guidelines, "AI-Assisted" means humans performed the lead vocal and primary instruments. In October, YouTube Music began rolling out labels for songs "created primarily using generative AI." In the first case, the key word is voluntary. An honor system works best when people are honest, which brings us back to the title of the Rolling Stone piece.

About that authenticity thing

Here's the irony. This is an industry that spent decades perfecting multitrack recording, synthesizers, drum machines, sampling, digital editing, and pitch correction. It built a whole economy on sounds that never existed in nature. Now it's worried about whether its music is authentic enough.

We're not saying the worry is silly. We're saying "authentic" has never meant "made without tools." Nobody demands a disclosure label for Auto-Tune. YouTube's partner guidelines, as TechCrunch reports, actually file pitch correction under "No Gen AI."

The question we think matters

We should also be honest about what the evidence does and doesn't show. It supports widespread, often undisclosed AI assistance. It does not show that most hit songs are now written or sung by machines. "Rubberz" made headlines because it's unusual.

So we think "Was AI used?" is turning into the wrong question. A better one is: what did the human actually contribute? Writing the melody? Singing the lead? Making the hundred small choices that turn a demo into a song? Or typing "80s synthwave, rapper lifestyle" and hitting enter?

There's a second question, and it's a big one: consent and fair pay for the artists whose work trained these systems. As Rolling Stone notes, those copyright questions remain "mostly unresolved." Whatever you think about AI as a tool, the people whose songs taught it deserve a seat at the table. Or as Lewis puts it, "if you're not at the table during those conversations of how it gets split, you're on the menu."

If you want to understand AI in creative work, start with the Rolling Stone investigation. It's rare to hear people this close to the work speak this plainly.