As AI music becomes increasingly sophisticated, how can listeners tell the difference? In a recent blind test commissioned by music streaming app Deezer, 97 percent of listeners could not tell AI-generated music from human-made music.
Seasoned musicians and producers may still be able to notice tell-tale signs. “There’s a certain sound quality associated with an AI song that gives it away. The musical arrangement tends to be repetitive, and it follows a very precise, mathematical pattern,” says DevM, a singer, producer and rapper. He shot to fame in 2019 with his collaboration with rapper Warboy as the duo won the Red Bull Spotlight, a hip-hop talent hunt.
Reena Gilbert, a music producer and mixing engineer, adds that because algorithms shape much of what we consume today, AI-generated music often has a similar tone, as it draws from existing content, making it easier for a trained ear to identify. Reena should know. Well versed with the technical and creative aspects of music, she started her career in the studios of Mumbai where she worked as live sound engineer for A. R. Rahman. Over the last 14 years or so, she has also evolved into an independent singer and composer with tracks such as “Berang” and “Jaane Kyun”.
It’s not always black and white, though. “It’s getting tougher to differentiate,” admits Raghav Meattle, singer-songwriter and founder of first.wav, a music company, as he cites the example of a demo a young Bengaluru-based musician sent him recently. “I had to listen to it multiple times, because something seemed off. The sound was too polished and refined. On further analysing the track and questioning the musicians behind it, we realised the song was AI-generated,” he adds.
Even a trained ear can be fooled sometimes. This is where audio detection tools come in. Several tools including Deezer’s AI Music Detector, Detect.music and letssubmit analyse uploaded audio files or streaming links to check if AI has been used in their creation. These detection tools are, however, not foolproof yet, but their results are best treated as indicators of AI involvement in a song rather than a definitive verdict.
Streaming platforms and AI music generators are taking cognisance of this issue and building systems that ensure songs are appropriately tagged at the submission stage. Speaking at the 10th edition of All About Music in Mumbai in August, Sourabh Pateriya, CEO of Soundverse, stresses the need for “limiting fully AI-generated content.” As surprising as it sounds, coming from the head of one of India’s primary AI music generators, he firmly believes AI-generated music can have repercussions, especially in terms of copyrights, ownership, artistic integrity and authenticity.
“We are figuring out ways to limit the number of tracks that can be generated per user, per hour, per day, per month. Whenever a track comes out of the Soundverse ecosystem, we add a very clear audio watermark to it,” he says. “If someone creates music on our platform and uses AI to generate a beat to layer the track, our automatic tagging system will attribute it to the person who made the song. The sound and underlying instrumental tracks will be labelled as AI-generated. We will pass this information directly to the DSP (Digital Service Provider), so the onus is not on the user. This is what we are currently building.” Apart from Soundverse, there are currently no other India-specific systems and tools that can detect AI music.
In August, Spotify announced it would start tagging non-human artiste profiles with an “AI Persona” badge. Spotify will also remove songs created by these AI personas from any algorithmic and editorial recommendations. The music streaming service will soon allow listeners to report profiles they believe are AI-generated. The team will review the reported profiles before sending the artists a notification to self-disclose or appeal. This feature will be rolled out in the coming months.
The age of Indian AI music
Not just around the globe, but in India too producers and musicians are already experimenting with AI and figuring out where this technology fits within their creative process. Take the viral “Chai Ki Tapri Pe” song from Samay Raina’s popular show “India’s Got Latent”, for instance. Created on Soundverse (complete with an AI-generated video) by producer and audio engineer Utkarsh Godbole, the song garnered over 500 million views on streaming platforms, already becoming one of the year’s biggest tracks.
According to Pateriya, who is a musician himself, AI helps democratise the music creation process. “I think the biggest value I see is that anyone who has a creative vision should be able to bring it to life with ease. They shouldn’t be lost in the execution process,” he said.
There is, however, a distinction between AI-generated music and music where AI is used as an assistive tool in a human-led process. DevM leans towards the latter. “For instance, I wanted to add a konnakol track to a song I was making and tried out a sample on Suno to understand what it would sound like,” he explains.
For musicians like DevM, AI tools are increasingly helpful in conceptualising a song, testing arrangements and adding layers to an existing track, whether it’s a guitar riff or a percussion rhythm. This helps them zero in on the ideal collaboration in the initial stages of song creation, before an artist is even approached.
Established mainstream musicians too haven’t shied away from experimenting with AI tools. A few months ago, Bollywood singer Sonu Nigam sang an AI-powered duet with the late singer Mohammad Rafi, at a live concert. Music maestro A.R.Rahman teamed up with Google Cloud in 2024 to create the world’s first AI music band called Secret Mountain. That same year, he used AI to recreate the voices of late singers, Bamba Bakya and Shahul Hameed (with family permissions), for a song in the Tamil movie Lal Salaam starring Rajinikanth.
With the deluge of AI music flooding streaming platforms, accurate detection and clear labelling become increasingly important. As the music industry’s AI policies continue to evolve, musicians and producers are sure of one thing: “Music creation is an organic process, and while technology can be a tool, it cannot replace human creativity. Let’s not forget the whole point of making art,” sums up Gilbert.
