Think you can spot an AI-generated song? This survey says you may be wrong

Songs generated by artificial intelligence (AI) tools are increasingly appearing on streaming services, making it difficult for users to distinguish them from those created by humans, a new survey reveals.

As many as 96% of respondents failed to tell which of three songs they just heard were entirely created by AI, according to a new Deezer/Ipsos poll that surveyed 2,000 American adults and was shared with Cybernews.

The findings highlight the difficulty of identifying AI-created music, which is now flooding streaming services.

Spotify has removed over 75 million spammy AI-generated tracks from its platform in the past year, in an effort to combat the degradation of listeners’ music experience caused by advancements in generative AI technology.

The crackdown on AI-generated songs came months after the platform angered listeners by promoting the Velvet Sundown, an AI-generated band whose songs were allegedly a mash-up of various other artists.

Some are excited about AI’s ability to create music that is comparable to that of human artists.

Nearly half (46%) of respondents in the Deezer/Ipsos poll said they felt “enthusiastic” about the similarity between AI-generated and human-created music, compared to 24% who were worried.

However, 81% American adults think 100% AI-created music should always be clearly labeled to listeners. Nearly seven in ten (69%) say AI-generated music threatens the livelihood of human musicians, artists, and composers.

Billie Eilish, Nicki Minaj, and Katy Perry, along with over 200 artists, wrote a public letter last year calling on tech firms not to develop AI music-generation tools that replace human artistry or deny artists fair compensation for their work.

In February, more than 1,000 UK musicians, including Annie Lennox and Kate Bush, released a silent album in protest against planned changes to copyright law.

The artists said the legislation would make it easier for AI companies to train their models using copyrighted work, potentially impacting livelihoods.

Unlock more exclusive Cybernews content on YouTube.