Sienna Rose has achieved a level of breakout stardom envied by musicians everywhere. Over the last four months, she’s developed a fan base of more than 4 million monthly Spotify listeners. Her star has risen in tandem with that of Olivia Dean, a fellow neo-soul vocalist who took home this year’s Grammy for best new artist. Rose’s Spotify biography extols her as “not just a performer, but a storyteller of the heart,” singing every word with a “sense of truth and beauty.”
But Rose didn’t attend the Grammys earlier this month because she’s neither a neo-soul artist nor the possessor of a soul to begin with. She is AI, top to bottom, an algorithmic interpretation of real artists in the genre it was built to exploit and ultimately replace. And most listeners have no idea.
AI-generated songs now make up nearly 40% of music uploaded to streaming services daily, according to a recent report. These songs are becoming more impressive by the day, too; in one study, only 3% of respondents could reliably distinguish between AI-generated and human-made music. And as of late January, six of Spotify’s top 50 trending songs in the U.S. were fully AI-generated.
We’re barreling toward a future where AI-generated music is the soundtrack to our lives, and most don’t even realize it. At a minimum, listeners deserve transparency. Just as Spotify and other major music platforms label songs they judge as “explicit,” they should clearly label AI-generated songs and artists to inform listeners who might not be able to make the distinction.
Rose is apparently the creation of Nostalgic Records, an AI-only label behind several successful AI-generated artists. If Rose continues its level of popularity, Nostalgic will make more than $1 million off its music this year. Meanwhile, Suno, the dominant generative AI-music platform, continues to see explosive growth, adding millions of active users and garnering a multibillion-dollar valuation.
To be fair, Suno is used for many worthwhile purposes. It’s admittedly fun to generate an EDM-country anthem for your sister’s birthday. As an artist and producer in L.A., I know many who already use Suno elements — a snare drum here, a synth stab there — in popular songs. AI-generated elements will be woven into the human music industry for the foreseeable future. But not if AI-generated artists destroy the industry first.
For every teen using Suno to ask their crush to prom, and every producer using it to craft a slinky drum groove, there are those bad actors exploiting Suno as a money grab. Multiple YouTube tutorials teach viewers how to use these tools to generate thousands of dollars in just a few minutes with “no musical talent, no equipment, and no studio” required. A handful of companies, including Nostalgic Records and Fifty Five Music Group, have begun to master these instruments.
New AI tools continue to upend a music industry already disrupted by the advent of digital music in the early 2000s. Digital production software and streaming services allowed independent artists to create and distribute music outside of traditional record labels. The resulting musical deluge has brought more songs to more people, but also made it more difficult for artists to break through. Since Spotify revenue is generated based on subscribers, AI artists are slowly eating more of the listening-share pie. Their royalty payouts come directly at the expense of human artists. A recent study estimated that AI artists will snatch away nearly 25% of human creators’ revenues by 2028.
Some might argue the negative impacts of AI music end there. It’s tough for music creators, but if listeners can’t tell the difference, why should they care?
For one, fully AI-generated music destroys the form’s artistic significance. Suno’s explicit goal is to solve the most arduous part of music: actually making it. Interacting with Suno is like interacting with ChatGPT, except instead of chat replies, the outputs are audio files. And though you can write your own lyrics, Suno is more than happy to generate them for you. “It’s not really enjoyable to make music now,” says Suno CEO Mikey Shulman. “People enjoy using Suno.”
But Shulman misunderstands the creative process. The friction of making art is a crucial part of what makes that art significant. Leonard Cohen spent a tortured five years writing and rewriting his song “Hallelujah.” If he had instead generated the final, mastered version in seconds, it’s hard to imagine it would have achieved the texture of turmoil, desire and grief for which the song is so known and beloved.
Frictionless art is also inherently derivative. Music, like broader American pop culture, has been criticized for creative stagnation. Generative music algorithms, trained on our existing musical heritage, would only exacerbate this trend with imitative slop. Our best chance of hearing truly innovative music in the coming decades is to make sure AI doesn’t crowd out real artists.
So far, the majority of AI-generated music isn’t great. Its cheesy lyrics, canned melodies and repetitive song structures can’t beat human music by quality alone. But the one thing humans cannot compete with is infinity. Sienna Rose has released 37 songs since September, a literally inhumane rate of productivity. Even the famously prolific and well-resourced Taylor Swift took two years to release roughly the same number of songs across her last two studio albums.
Nothing is stopping companies from flooding streaming services with millions, even billions, of AI-generated songs and artists. Suno currently generates 7 million songs per day. If every one was uploaded to Spotify, within a year human-made music would constitute less than 4% of the platform’s music.
Few listeners would choose this future. Even employees at Suno who’ve confided in me personally agree. Yet we’re already spiraling into this world — unless we take action. Tagging songs with a prominent “AI” label, similar to the “explicit” label, is a neat and necessary first step toward stopping the spiral. The streaming platform Deezer announced in June that it would publicly tag fully AI-generated songs and exclude them from algorithmic recommendations. The indie music platform Bandcamp has decided to ban AI-generated music entirely.
To its credit, Spotify has announced steps towards transparency. But their promises are vague at best. The company wants to rely on artists and record labels to self-disclose whether a song is AI-generated, and plans to list those disclosures in the song credits, which are well out of view for Spotify users. Moreover, the company announced these changes in September, but there’s still no evidence of AI labels anywhere on the platform. Apple Music has yet to take any stance at all.
Spotify’s mission statement since about 2018 is “to unlock the potential of human creativity — by giving a million creative artists the opportunity to live off their art and billions of fans the opportunity to enjoy and be inspired by it.” If Executive Chairman Daniel Ek is serious about this, they must act; or their users will. We can switch to streaming services, like Deezer, that care more about human creators, and sign online petitions to unite behind human artists.
Human music is as old as human language, and perhaps even older. As such, it belongs to humans, not corporations and their robots. Inform us of the distinction so we can make the choice ourselves.
PJ Frantz is a Los Angeles–based music artist and producer focused on the intersection of music, technology and culture.
