YouTube is Testing Prompt-Based Feed Customization to Break the Recommendation Loop
Table of Contents
Beyond the Algorithm: Directing Your Own Discovery
For years, the YouTube experience has been defined by the ‘black box’ of the recommendation algorithm—a system that observes your behavior and feeds you more of the same. While effective for retention, it often creates an echo chamber or a ‘filter bubble’ where users struggle to find new content without manually searching for it. Google is now attempting to give users a steering wheel with the introduction of a new AI-powered feature called “Your custom feed.”
Rolling out currently to a subset of users in the U.S., the feature manifests as a chip at the top of the YouTube Home page. Unlike the static category chips (such as ‘Music’ or ‘Gaming’) that filter existing tags, this tool allows users to enter natural language prompts to generate a bespoke feed of videos based on specific, real-time intent.
The utility here isn’t just about finding a specific topic, but about defining a mood or a constraint. Google has highlighted use cases such as requesting “something different beyond my usual feed” to break a content rut, or more granular requests like “guided meditations under 10 minutes to unwind after work.” By shifting the power from passive observation to active prompting, YouTube is essentially treating the homepage like a generative AI interface.
The Privacy Trade-off: Why History Matters
The feature is not available to everyone by default, and there is a significant technical requirement: users must have both their YouTube search and watch history enabled. This requirement underscores the hybrid nature of the tool. While the prompt provides the direction, the underlying AI still relies on the user’s historical data to calibrate what “something different” or “relaxing” actually means for that specific individual.
For those who have opted for a more private experience by pausing their history, the “Your custom feed” chip will likely remain invisible. To enable these settings, users must navigate to the “Your data in YouTube” section within their profile settings. This creates a tension common in Google’s current product strategy—offering high-value AI utility in exchange for deeper data telemetry.
Contextualizing the Shift in User Agency
This move comes as Google aggressively integrates Gemini-powered capabilities across its ecosystem. In the context of YouTube, this is a strategic pivot. For a decade, the goal was to predict what the user wanted before they knew they wanted it. However, as users become more accustomed to prompting AI (via ChatGPT or Perplexity), the expectation is shifting toward active curation.
By allowing users to “tell us in your own words” what they want to see, YouTube is acknowledging that its predictive algorithms occasionally fail or become too repetitive. This is a subtle but important shift in the power dynamic between the platform and the viewer. Instead of the algorithm deciding your evening’s entertainment, you are now providing the creative brief.
How to Access and Use the Custom Feed
If the feature has reached your account, the process is straightforward:
- Locate the “Your custom feed” chip at the top of the Home screen on mobile or desktop.
- Enter a specific prompt in the text box. You can be as vague (e.g., “Surprise me”) or as detailed (e.g., “Deep dives into 90s synthesizers”) as you wish.
- The resulting feed refreshes automatically, though users can edit their prompt at any time to pivot the content stream.
While currently limited to English-language users in the U.S., it is expected that this will expand globally as Google refines the natural language processing required to map complex prompts to the massive library of YouTube’s video metadata. The only lingering question for power users is whether this will remain a free utility or eventually be bundled into the YouTube Premium subscription tier, a move Google has used previously to gate-keep experimental features.
