Spotify’s AI-Powered DJ Gains Traction, Expands Into European Markets

As terrestrial radio becomes increasingly irrelevant to younger listeners, Spotify’s new AI-powered radio feature, “DJ,” is breathing new life into the medium.

Featuring curated playlists with audio commentary around the tracks, DJ was first released by Spotify in February for U.S. and Canadian users. The experience is personalized for each user, unlike broadcast radio. The audio commentary, however, is generated through artificial intelligence using Sonantic, a voice engine acquired by Spotify last year.

Now, just a few months later, Spotify has expanded the feature to Spotify Premium users in the UK and Ireland. The feature is now available for users in those markets on both iOS and Android devices, which can be accessed through the Music Feed by tapping on the DJ card. According to Spotify, the decision to expand to those markets was influenced by social media engagement from UK and Irish users.

Playing a crucial role in the development of “DJ” is Xavier “X” Jernigan, Spotify’s head of cultural partnerships. Not only is DJ’s voice modeled after X, but he works with Spotify’s team of music editors, scriptwriters, and cultural experts to ensure DJ is delivering the most relevant information to users, adding a human element when training the AI.

Upon my first use of Spotify’s DJ feature, X identified itself as an AI, a conscious decision made by Spotify’s team, but clarified he doesn’t set times or switch on your lights, hitting the post into the intro of “Disparate Youth” by Santigold, my go-to anthem of 2012.

“We have cultural experts and creative producers in-house. We use all of that knowledge of culture and what’s happening in music, our relationship with artists and our relationships with labels to help in the curation process,” he said. “It’s not a replacement for a DJ, it’s an enhancement.”

Jernigan is no stranger to Spotify listeners. He was a former host of the company’s first foray into DJ-led programming, “The Get Up.” He says the show played a crucial role in developing the voice for DJ, considering he had no formal radio hosting experience.

“When ‘The Get Up’ came, I just had to be unapologetically myself. And that all ties back to DJ. What you hear is me. You don’t hear anything different,” he said.

Jernigan emphasizes that listeners tune in to hear the host’s thoughts and delivery of headlines, not just the headlines themselves. He sees the AI-powered DJ as a knowledgeable, relatable friend, sharing music and culture with the listener.

Spotify’s technology creates a realistic listening experience, with Jernigan praising the voice AI technology for sounding just like him and capturing the energy and connection with listeners.

“You said it sounds just like me, my mama thinks so too. When I played it for her, she didn’t realize it was AI and text-to-speech, she thought it was me,” he said.

Recording the voice database was an intensive process. While Jernigan did not disclose how long the process took to train the voice model, the process consisted of Jernigan recording lines of script to train the model, focusing on creating a conversational tone and connection with the listener.

“We’re not competing with radio, because this is personalized. Radio is broadcast,” he said. “This is a personalization play for each and every listener.”

Emily Galloway, Spotify’s Head of Product Design for Personalization, explained that the platform has been using AI and machine learning for personalization and driving discovery for many years, with Discover Weekly being their first successful algorithmically created playlist in 2015.

The decision to add an AI-powered feature was driven by the fact that context is powerful in driving discovery. Since launching DJ, Spotify has found that when listeners hear commentary, they are more likely to try something new and are more willing to try a song they may have otherwise skipped.

“Context is a really powerful thing. If people know why they are being recommended music, they’re more likely to listen to it,” Galloway said.

Spotify wanted its AI voice to be approachable, human, and friendly, hence the selection of Jernigan, who was well-received by listeners of “The Get Up.” The combination of Spotify’s recommendation algorithms with Jernigan’s voice model is proving to be a successful enhancement to the listening experience. In fact, some Spotify users inquired through Twitter about marrying X.

User engagement and discovery have been positive since DJ’s launch. Fans who come back listen to DJ 25% of the time, and more than half of first-time listeners return the next day. Millennials and Gen Z make up 87% of its users, which is particularly compelling considering these groups are turning away from conventional media.

“We didn’t design a voice, we designed a persona,” said Galloway. “We really are driving towards making these more meaningful connections between listeners and creators, listeners and artists.”