YouTube Lets Users Prompt AI for Custom Homepage Feeds

YouTube is rolling out a new feature that lets users generate personalized video feeds through simple text prompts, marking another step in platforms’ efforts to hand more curation power to individuals rather than relying solely on opaque recommendation algorithms. The tool, called “Your custom feed,” is now available to signed-in users in the United States on both the mobile app and desktop, with support limited to English for the initial launch.

Users access the feature by selecting the “Your custom feed” tab at the top of the homepage and typing a description of the content they want. Prompts can specify interests, time constraints, or even mood, such as short guided meditations or in-depth podcasts on artificial intelligence applications. The resulting feed can then be pinned for repeated access. Prompts remain editable, allowing users to refine or replace them without starting over. To enable the tab, viewers must keep their search and watch history active in account settings, a requirement that underscores the feature’s dependence on existing data collection practices.

The move follows similar experiments elsewhere. Spotify has offered prompted playlists for music and podcasts, while Instagram has expanded topic-based controls over its Reels recommendations. YouTube’s version stands out for its focus on video and its integration directly into the homepage experience, where most users begin their sessions. Company statements emphasize flexibility, noting that the system can generate entirely new collections when prompts change.

Early details also include modest feedback mechanisms. A three-dot menu on the custom feed tab allows users to flag problems if the results diverge from expectations. The feature arrives shortly after YouTube began labeling certain AI-generated videos, suggesting the company is simultaneously expanding and attempting to manage new generative tools.

For observers tracking how recommendation systems shape attention, the change raises questions about the balance between user agency and platform control. While prompts give viewers a direct way to steer results, the underlying model still determines which videos qualify and in what order. History requirements mean the feature operates within the same data environment that powers standard recommendations, potentially limiting its ability to break established patterns of viewing. Over time, widespread adoption could shift how creators reach audiences, favoring those who align closely with prompt-friendly topics or formats.

The rollout remains geographically narrow for now, leaving open how or when similar options might appear elsewhere. As other services test comparable interfaces, YouTube’s version will likely serve as a test case for whether descriptive prompts meaningfully improve discovery or simply add another layer to already complex personalization systems.