AI in Music: Taking the Soul Out of the Song

As AI rapidly becomes more accessible and prevalent in the music industry, musicians and journalists are investigating the various ways in which AI is affecting music creation and streaming.  

AI can be used to generate music or generate fake streams on platforms such as Spotify, Apple Music, YouTube Music, and Amazon Music. According to Wired, in March of 2023, an AI generated song that fakes the voices of artists Drake and The Weeknd went viral but was eventually removed from Spotify.  

More recently, on April 19, 2024, Drake himself used AI to replicate the voices of hip-hop artists Snoop Dogg and the late 2Pac. “Taylor Made Freestyle”, the song featuring the AI verses, was  met with severe backlash from hip-hop fans as well as 2Pac’s estate ran on behalf of his family. The estate allegedly ordered a cease-and-desist against Drake, referring to the use of 2Pac’s  voice in AI as a “blatant abuse of his legacy.” 

Due to the recent uprise in the music industry's use of AI, up-and-coming artists express concern about how AI may affect their careers. 

Stream Farming  

Grace Adair, a 19-year-old singer/songwriter from Bothell, WA, releasing music under the pseudonym Graceey, said that AI can be used to fake streams, noting Spotify specifically as it is  the primary website that she releases music on.  

“After I released Quarters, my first ever single, Spotify allows you to see the real-time streams on it,” said Adair. “For example, I hit ten thousand streams, and then I looked the day after and saw that my streams went down by one thousand.”  

Adair said that she was initially confused as to why her streams dropped so significantly, until doing research and learning of an AI streaming method referred to as “botting.”  

According to a study by France’s Centre National de la Musique, this method is also referred to as “stream farming”, wherein AI bots are used to create mass amounts of streams.  

“[Botting/stream farming] can be done a number of ways,” Adair said. “Sometimes, artists who try to be cheeky, because you make money per stream on a song, will try to use AI to bot their song, to have it on repeat and to create mass playlists, so that it has an increase in streams.”  

However, Adair said that Spotify can typically notice when an actual human is manipulating streams by leaving a song on repeat, but with AI, users can create multiple accounts and playlists that each appear unique to boost an artists’ streams without raising Spotify’s suspicions. 

“We have been seeing more [botting] in this music industry because of how new we are to AI, and because we don’t really have any regulations in terms of the legality of it. It’s a lot easier to get away with botting streams,” Adair said.  

Adair said that people can use AI to create an algorithm of fake accounts that are each unique and can create these accounts rapidly, making it difficult for Spotify to properly assess where the  accounts and streams are coming from.  

There is also a financial aspect to stream farming. According to the article “Spotify Farm: Uncovering the World of Streaming Farms” by copywriter Sam Carr, both artists and Spotify itself are losing money at the expense of fake streams.  

In an experiment by Vice journalist William Bedell, Bedell created a streaming farm to create fake listens on songs he uploaded to Spotify. Carr reported that Bedell’s generated revenue was about $30 a day from the fraudulent streams, which Spotify was paying him for. 

According to Carr, Spotify’s lack of AI-detection systems has cost the company thousands of dollars in paying an artist for fake streams. Faux streams put not only Spotify’s money at risk, but its advertisers as well.  

How AI Affects Aspiring Musicians and their Careers  

Botted streams and AI-powered mass streaming farms pose a potential threat to careers in music  – the lack of a true fanbase.  

Adair said that the biggest reason why streams are so important – on Spotify in particular – is that having many streams indicates that the musician earning said streams is increasing the app’s user base. Spotify makes note of that, then incentivizes musicians by paying them per stream.

“If you have a certain number of streamers returning, Spotify’s going to go, ‘hey, I love that,  you’re obviously bringing people back onto the app and helping them stay on it, so we’re going  to reward you by paying you money to do that,” said Adair. “That’s the most important thing  Spotify wants. It’s not a platform, it’s a music library where they can compile all this music, and your job as a musician is to keep bringing people onto their app so they’ll want to buy Spotify  premium, and they’ll keep coming back to your music.”  

Adair said that because of this, botted streams will affect the authenticity of these incoming  streams, providing an example of an artist that has a song with 1,000 streams who might think that is the exact number of fans they have, when many of the streams are AI.  

“They’ll think they have that many fans, until they come to find out that they’re all botted fans,  and they’re getting ready to go on tour because Spotify’s like, “oh, you’re really a successful musician!’” Adair said. “They go on tour, [and] they’re not going to make any money from that because they don’t have real fans.”  

How AI Affects Music Creation  

AI is not only used to stream music but can also be used to create music. Whether vocals or instrumentals, there are various AI programs that allow users to create entire songs, often  mimicking the voices and styles of existing, popular artists.  

Adair said that there are several websites and programs that allow its users to take the voice of a singer — naming Taylor Swift as an example — by taking snippets of the artist’s songs, then creating a song based on what the AI knows about Swift’s music.  

“[AI is] still based off another person’s work, so you can’t just feed AI and say, ‘Okay, I want to write a song about this.’ No. It’s going to use a chord structure, it’s going to use a beat and it’s going to use synth from an already existing song,” Adair said.  

Adair relates the process of creating AI music to the rising controversy regarding AI-generated art, stating that although the art may be an AI-generated piece, it is still built on existing, human made work.  

Carl Pedida, a music producer and vocalist, said that artists should only use AI to help with their song ideas, rather than letting AI produce everything for them.  

“So, say you have a shaggy prompt or whatever and [the AI] gives you an answer on how to structure your song, or something like that,” said Pedida. “I feel like that is helpful in a sense, but  that’s very like... you’re only kind of using [AI] as another tool, but using it to create the entire  song... I feel like it ruins a lot of the creativity.”  

Pedida said that he advises beginner musicians to stay away from AI and instead encourages them to explore their own creativity from early on so that they do not become reliant on AI  programs.  

“I feel like having access to AI, being absolutely transparent, makes creativity very lazy,” Pedida said. “If you’re a beginner producer who is just starting out and trying to get accustomed to being creative and exploring what your brain is capable of and then going to AI whenever you can’t figure something out, it kind of stunts your creative growth.”