Spotify’s AI Chatbot Turns Music Discovery Into a Conversation

Spotify is testing an AI chatbot that lets Premium subscribers request music, podcasts, and audiobooks through text or voice. The feature can use personal listening information—including playlists, favorite artists, and history—to respond to requests that are difficult to express through filters or a traditional search box.

A listener might ask for unfamiliar music with the energy of a favorite running playlist, request a calmer transition without changing genres, or ask when they first encountered a particular song. That is a more flexible interaction model than choosing from mood tiles or accepting a generated playlist with little context.

Conversation can capture intent

Music discovery often begins with incomplete language. People remember a feeling, a setting, an instrument, or a relationship between artists rather than an exact title. A conversational system can turn those fragments into a sequence of clarifying questions.

This could be especially useful across Spotify’s expanding catalog. Someone looking for a podcast episode about a specific idea or an audiobook for a particular trip may benefit from a system that understands constraints across format, length, tone, and prior listening.

Personalization can become too comfortable

The same listening history that makes a response relevant can also narrow it. Recommendation systems already tend to reinforce familiar patterns because a predictable click is easier to optimize than a meaningful surprise. A chatbot may make that feedback loop feel more intentional even when the underlying recommendation remains conservative.

Spotify should therefore make room for explicit discovery controls: less familiar, outside my usual region, from independent labels, released before a certain year, or unlike what I normally play. The interface should also explain why an item was recommended and allow listeners to correct assumptions.

The assistant should serve listening, not replace it

Conversation is valuable when it reduces friction between intention and audio. It becomes distracting when the interface encourages people to talk about listening more than they actually listen.

The strongest version of this feature will disappear quickly after it has helped. It will create a queue, preserve the reasoning behind it, and let the music take over. The success metric should not be chatbot messages. It should be whether listeners discover work they value and understand how their choices shaped the result.