Spotify opens Taste Profile to US Premium listeners

Spotify began rolling out its AI-driven "Taste Profile" to Premium subscribers in the United States on September 23, 2026, allowing users to see and directly edit the data that powers the platform's recommendation engine. The feature, first demonstrated at SXSW in March, had previously been limited to a New Zealand beta. By expanding to its largest market, Spotify is testing whether giving listeners granular control improves engagement and reduces the "filter bubble" effect that often leads to overly homogeneous suggestions.

How Taste Profile works under the hood

Taste Profile taps into the same machine-learning pipelines that generate Discover Weekly, Made for You mixes, and the annual Spotify Wrapped recap. While the exact model architecture is proprietary, Spotify has confirmed it relies on a combination of collaborative filtering and deep-learning embeddings that map users, tracks, podcasts, and audiobooks into a shared latent space. When a user submits a natural-language command—e.g., "show me more upbeat indie rock"—the system translates the request into vector adjustments, nudging the user's position in that space. The updated vector is then used by downstream ranking models to re-score candidate items for the Home feed.

User workflow and latency

To access the tool, users tap their profile picture in the mobile app, scroll to the "Taste Profile" entry, and review a visual breakdown of their current taste across three categories: music, podcasts, and audiobooks. A "Tell us more" field accepts free-form text; after submission, Spotify processes the request on its cloud inference servers. According to the rollout notes, the revised recommendations appear within a few hours, indicating batch-update cycles rather than real-time inference.

Immediate impact on core recommendation products

Because Taste Profile directly influences the latent user vector, any adjustment propagates to all downstream recommendation products that share the same embedding. This includes:

  • Discover Weekly – the weekly curated playlist that already reflects long-term listening patterns.
  • Home feed – the constantly refreshed carousel of songs, podcasts, and playlists.
  • Spotify Wrapped – the year-end summary that aggregates listening statistics and highlights.

Early internal testing, referenced in the launch announcement, suggests that users who request broader genre diversity see a 12% increase in unique track plays within the first week, though Spotify has not released full metrics.

Why Spotify is betting on user-controlled AI

The music streaming market is increasingly competitive, with Apple Music, Amazon Music, and emerging AI-first platforms vying for listener attention. Spotify's algorithmic edge has been a key differentiator, but critics have noted that its recommendations can become stale, especially for power users who exhaust the platform's suggested catalog. By exposing a controlled interface to the recommendation engine, Spotify aims to:

  1. Reduce churn – giving users a sense of agency may keep them subscribed longer.
  2. Mitigate echo chambers – explicit requests for variety can counteract over-personalization.
  3. Gather fine-grained feedback – the natural-language inputs provide labeled data that can improve future model training.

Risks and open questions

While the feature promises more personalized discovery, it also raises concerns:

  • Algorithmic transparency – users see a high-level view of their taste vector but not the underlying weighting scheme, leaving room for misunderstanding.
  • Data privacy – the system stores user-generated text alongside listening histories; Spotify has not detailed how this data is protected or whether it is used for advertising.
  • Bias amplification – if many users request similar genre shifts, the model could over-fit to trending preferences, potentially marginalizing niche content.

Spotify has not disclosed a roadmap for expanding Taste Profile beyond the US or for lowering the age threshold. The feature remains in beta, and the company has not announced a timeline for a global rollout.

Industry reaction and comparable efforts

Analysts note that Spotify is the first major streaming service to let users edit the recommendation engine directly. Apple Music recently introduced a "Play Like" toggle that lets users mimic another listener’s taste, but it does not expose the underlying model. Meanwhile, podcast-focused platforms such as Stitcher are experimenting with user-driven recommendation filters, though without AI-level adjustments.

The move aligns with a broader trend of giving end-users more control over AI outputs, echoing recent debates around transparency in recommendation systems. For a deeper look at how AI-driven personalization is reshaping media, see the coverage of the TechCrunch Disrupt 2026 Offers $200 Ticket Discount as AI Leaders Converge.

What to watch next

Spotify has not indicated when Taste Profile will exit beta or whether it will be extended to free tier users. Key indicators to monitor include:

  • Adoption metrics – the proportion of US Premium users who engage with the feature within the first month.
  • Engagement shifts – changes in daily active users, session length, and cross-category consumption (music vs. podcasts).
  • Regulatory scrutiny – as EU and US lawmakers examine algorithmic transparency, Spotify may need to disclose more about how user edits affect the recommendation pipeline.

If the beta proves successful, we could see a cascade of similar controls across other recommendation-heavy services, potentially reshaping the balance between algorithmic curation and user intent.

How does Taste Profile differ from existing Spotify controls?

Taste Profile adds a visual summary of the AI's perception of a user's taste and a natural-language input field that directly modifies the underlying recommendation vectors. Prior controls only let users hide specific tracks or toggle a "kids music" filter, which did not affect the core recommendation model.

Will Taste Profile affect Spotify Wrapped?

Because Wrapped is generated from the same user embeddings that power other recommendation products, any changes made through Taste Profile will be reflected in the year-end summary, altering genre and artist statistics shown to the user.

Corroborating reports: The launch was also covered by Ars Technica, which highlighted the broader industry shift toward user-editable AI models Ars Technica.

Related coverage

Explore more on this topic