As the holiday season reaches its peak, a new wave of Christmas music is flooding streaming services – but not all of the artists are who they seem. A growing number of tracks credited to seemingly unknown musicians are actually generated by artificial intelligence, raising questions about authenticity and copyright in the music industry. This practice is prompting concern among artists and industry experts about the potential devaluation of human creativity and the murky legal waters surrounding AI-created content.
-
Jacobien van der Kleij
Online Editor
-
Jacobien van der Kleij
Online Editor
Dean Snowfield, Sleighbelle, and Daniel & The Holly Jollies: they appear to be just ordinary artists releasing Christmas songs on Spotify. But in reality, these artists don’t exist, and their Christmas covers were created using artificial intelligence (AI). SoundCloud and YouTube are also flooded with similar tracks.
The influence of AI on the music world grows every year, says music expert Kris Keijser. He himself produces tracks with artificial intelligence, such as Zo Zomer by ‘John de Koning’ from 2024. “There’s a kind of noise in it; it just doesn’t sound good. If you were to make it now with the AI app Suno, you’d really hear the leap in quality. The technology has improved, but you can still tell it’s fake.”
30,000 AI Tracks Per Day
Spotify hasn’t disclosed what percentage of its total library consists of AI tracks. However, the streaming service removed 75 million “spam tracks” last year. These are songs created by AI that sound almost identical to real artists.
Similar streaming services have data on the amount of AI content available. France’s Deezer, for example, discovered through its own tool that approximately one-third of all uploaded music is AI-generated—more than 30,000 tracks per day.
Generative AI learns from all the music available worldwide.
Data scientist Janne Spijkervet has conducted extensive research into artificial intelligence and, along with rapper Willie Wartaal, produced the song Abbus, the entry for the AI Songfestival in 2020. “The first signals that AI would become big came ten years ago. That manifested itself in playlists, which are suggested by AI.”
Late 2023 saw the release of Suno, one of the first AI programs that allows users to create songs themselves. “That’s when generative AI really took off; that technology learns from all the music available worldwide,” says Spijkervet. “Every AI-generated track is based on existing artists and songs. And that appeals to listeners.”
For example, Sleighbelle’s AI-artist cover of Santa Baby on Spotify has been played over 14 million times. The track appears in playlists alongside flesh-and-blood artists like Ella Fitzgerald, Michael Bublé, and Brenda Lee.
“It’s happening more and more often that certain artists are being imitated. Tim Knol recently experienced this,” says Spijkervet. “That’s harmful to artists, because it mocks their work.” Moreover, they are losing income.
No Permission
It’s also often difficult for listeners to tell the difference between AI-generated music and songs by real artists. A recent study by Deezer and Ipsos found that 97 percent of the 9,000 participants couldn’t identify which tracks were made by AI and which weren’t.
According to Spijkervet, it should be clearer to listeners when they are listening to AI content. “Copying or appropriating someone’s voice goes beyond covering a song. It’s also often not labeled as AI music on Spotify or YouTube,” she says.
AI Christmas songs often remain quite generic. A really good Christmas song has to touch you.
Spijkervet expects that AI music generators will eventually reach a quality plateau. This is due to the many lawsuits that record labels have filed against platforms like Suno and Udio for using existing music without permission. “At some point, these platforms will really have to comply with the regulations and won’t be able to use all the music to train their models.”
Music expert Keijser says there are benefits to using AI for making music. “It can inspire you as an artist and help you further. For example, if you have a certain text in your head and you know how you want it sung, the program comes up with something slightly different.”
Imperfections
Coming up with Christmas songs is more difficult, he says. “Making a really good Christmas song is hard: it has to touch you or there has to be a story behind it. AI Christmas songs often capture a generality.”
Moreover, you often hear that it sounds fake, especially when you listen closely to the voice, says Keijser. “In songs by The Band, The Beatles, or D’Angelo, there are certain imperfections that AI doesn’t make, and that’s precisely what makes them so good. There’s also a groove that you can’t replace with AI.”
Spijkervet adds that Christmas music is, above all, a genre where authenticity and genuineness are important. “It’s fun to experiment with AI and get creative with it, but Christmas songs that do well have existed for 40 years, and it’s always the same artists and songs. The Christmas genre draws on tradition, coziness, and nostalgia, and AI as an experiment doesn’t fit into that.”
