YouTube launches AI-powered custom feed builder with natural language prompts

YouTube’s Natural Language Feed Generation

According to The Verge AI, YouTube is rolling out an AI-powered custom feed feature that generates personalized video collections based on natural language descriptions. Users can request feeds around specific interests, moods, or topics—such as “guided meditations under 10 minutes” or “deep-dive tech podcasts about AI”—and pin them to their homepage for quick access. The feature is currently available to English-language users in the US who are signed in on YouTube’s mobile app or desktop interface.

The mechanism is straightforward: users click the “Your custom feed” tab at the top of the YouTube homepage and enter a prompt into an AI text box. YouTube’s system then curates a video collection matching the request. Prompts can be edited at any time to regenerate the feed, allowing users to refine their requests iteratively.

How This Positions YouTube Against Competitors

YouTube’s approach mirrors Spotify’s AI-powered playlist generation, which allows users to describe a mood or activity and receive a custom music collection. Instagram introduced similar algorithmic control in December 2025 with topic-list-based curation for its Reels feed, though that mechanism relies on explicit topic selection rather than descriptive prompts. YouTube’s natural language interface is arguably more flexible, letting users express nuanced preferences without predefined categories.

The feature requires users to have search and watch history enabled in their account settings to access the “Your custom feed” tab. If YouTube’s AI misinterprets a request, users can report issues through the 3-dot menu and select “Something wrong?” to submit feedback.

Why This Matters

This rollout signals YouTube’s pivot toward generative AI as a personalization layer rather than relying solely on algorithmic recommendations. For users fatigued by opaque recommendation systems, custom feeds offer agency—the ability to explicitly request content rather than passively receive algorithmic suggestions. However, success depends on the AI’s ability to accurately parse subjective requests and surface relevant videos; widespread mismatches could undermine adoption.

The limited initial rollout—English-language US users only—suggests YouTube is testing infrastructure and user feedback before broader deployment. The feature’s expansion to other regions and languages will determine whether it becomes a standard YouTube interface element or remains a niche tool for power users.