Spotify is launching Talk to Spotify, a beta feature for Premium subscribers that introduces a conversational interface inside the mobile app. Type or speak, and the AI assistant picks what to play, answers questions about your listening history (when you first heard a song, which genres you listen to most), and explores podcast and audiobook content. Available as of yesterday in the US, Ireland, and Sweden on iOS and Android, in English, for adults only.

Behind it is a mix of AI technologies (Spotify uses several depending on the task, without disclosing which), and the feature relies on your listening data: playlists, favorite artists, history. You can refine the selection with follow-up requests (“more recent,” “add this artist”), save tracks, add to queue.

Why it matters to you. It’s an example of an AI assistant in an established daily-use context, not a separate experiment. Amazon Music integrated Alexa Plus similarly last year, but Spotify goes beyond recommendations: it lets you interrogate your own listening history and control playback with natural language. It’s the pattern taking hold: the conversational interface as a second access point to data you already own, alongside traditional navigation.

If you want to try it. The feature appears in Home and Now Playing if you’re Premium and in the right market. Spotify warns it’s a beta and “responses won’t always be perfect.”

In detail

What came before

Spotify had already introduced AI features: the AI DJ that speaks with synthetic voice and suggests music, and Prompted Playlist, which builds playlists from text descriptions. Talk to Spotify surpasses them because it’s conversational: it doesn’t just generate a playlist from a prompt, but keeps the exchange open to refine, query, and explore.

Amazon Music moved in the same direction in 2025 with Alexa Plus integration, but without direct access to personal history with the same granularity.

How it works

The assistant appears as a text box (or microphone) in Home and Now Playing view. You can ask it to play “songs I’ve never heard,” then specify “add this artist” or “more recent.” Or query your history: “When did I first listen to this song?”, “Which genres have I listened to most in the past weeks?”.

You can also ask general questions about the content you’re listening to: release date of a track, other titles by the same audiobook author, previous episodes of a podcast. Answers draw from Spotify’s catalog metadata and your listening data.

Spotify states it uses “a mix of proprietary AI technologies and models from multiple vendors, choosing what works best for the task.” It doesn’t specify which models or how it balances privacy and personalization: listening data stays on Spotify infrastructure, but it’s unclear if and how it’s shared with model vendors.

Implications and limits

It’s a case of consumer AI assistant with repeated use: not a one-off experiment, but a second access point to your own data. The fact that Spotify is testing this on millions of Premium users signals the conversational interface is considered mature enough to sit alongside established graphical navigation.

Spotify itself cautions: “Responses won’t always be perfect.” There’s no public trace of which errors emerged in internal testing, or how they handle hallucinations on metadata or history. The feature stays in beta and will expand based on feedback.

The choice to limit to adults and Premium users signals caution (smaller attack surface, paying users more tolerant of issues), but also that Spotify sees this as an incentive to subscribe, not a baseline utility.

Where sources diverge

The Verge and TechCrunch agree on the main facts. TechCrunch adds the detail about multiple AI model vendors (confirmed by Spotify in direct response), while The Verge focuses on comparison with Amazon Music and positioning as a response to criticism of Spotify’s recommendation algorithm.

Neither source reports accuracy metrics, or how many users have beta access at launch. Spotify hasn’t shared plans for geographic or language expansion beyond English.