AI Music Shock: Can You Tell The Difference? 🎶🤯

Tech & Science

Streaming service Deezer recently conducted an experiment with research firm Ipsos, revealing a surprising finding: 97 percent of participants were unable to distinguish between fully AI-generated and human-made music. However, this result represents only part of the story. In the survey, 9,000 participants listened to three tracks and were asked to identify which, if any, were entirely AI-generated; those who failed to correctly identify all three tracks were categorized as “failures.” Consequently, according to Deezer and Ipsos, any participant who correctly identified two or three of the tracks was deemed unable to differentiate between AI-generated and human-created music.

Intrigued by the initial findings, I decided to conduct my own, less formal experiment. I presented the same three tracks to ten individuals, posing the same question. While a significant number of participants struggled to identify the AI-generated songs – only one person correctly identified all three – the results shifted dramatically when the responses were analyzed individually. Without bundling the responses, 43 percent of participants were able to accurately determine whether a track was AI or human-generated. Notably, several individuals commented that one of the songs was so poorly produced that they suspected it was intentionally AI-generated, leading them to incorrectly identify it as real. Deezer’s study participants were particularly taken aback, with 71 percent expressing surprise at the results and 51 percent indicating that they felt uncomfortable with the inability to distinguish between the two types of music.

Opinions were sharply divided regarding the impact of AI-generated art, with 51 percent believing that such creations would result in “more low-quality, generic sounding” music. Remarkably, only 40 percent indicated they would skip AI music entirely if they encountered it knowingly. Despite this hesitancy, a strong consensus emerged concerning the need for transparency: 80 percent of respondents called for clear labeling of AI-generated music. Deezer has taken the lead in this area, implementing a system capable of automatically detecting and labeling 100 percent of AI-generated content from popular models such as Suno and Udio. The platform also excludes music labeled as AI from its algorithmic recommendations. Spotify recently announced steps to address problematic AI content on its platform, primarily focusing on policies related to AI impersonation and a new spam filter designed to curtail abusive behavior. However, Spotify refrained from implementing blanket labeling, instead pursuing a standardized credits system, stating, “The industry needs a nuanced approach to AI transparency, not to be forced to classify every song as either ‘is AI’ or ‘not AI.’” This system would largely depend on labels and artists honestly disclosing the use of AI, even in instances like mixing. A key figure emphasized that this is “not a technical problem. It’s a transparency issue and it’s an ethical issue” that demands a comprehensive response from all stakeholders.

The challenge extends across the music ecosystem, from creators and distribution services like DistroKid to the streaming platforms themselves, demanding responsible action. Notably, the volume of AI-generated music being uploaded is substantial and growing rapidly; Deezer, for example, reports receiving over 50,000 such tracks daily, representing more than 34 percent of all music added to the service. According to CEO Florian Moussallam, despite this dizzying influx of AI content, it accounts for only 0.5 percent of streams, and the vast majority of that small percentage is identified as fraudulent. Acknowledging the volume as a challenge for Deezer, Moussallam maintains that it doesn’t fundamentally alter the user experience, asserting that “Humans continue to create music, and they will continue to listen to the music made by real artists.” Survey results reveal widespread concern, with 70 percent of respondents viewing fully AI songs as a threat to musicians’ livelihoods, and 64 percent believing AI could diminish overall creativity.