YouTube is reportedly testing a new feature that will allow users to refine the platform’s homepage recommendations with simple AI prompts. The new experimental tool is called “ Your Custom Feed .” It allows users to type in specific interests so that the algorithm can deliver more relevant suggestions.
YouTube tests a new AI tool that lets users refine homepage recommendations
With the new experimental tool, Google aims to give YouTube viewers more control over what appears on their homepage and may reshape how they discover new topics and channels across the platform. A few users have already started seeing the Your Custom Feed option. The tool appears as a chip at the top of the homepage.
Users can enter short prompts, such as “enjoying history podcasts” or “want more music videos,” and they can push the system toward content they want to explore. The algorithm will try to push the related recommendations even if the history does not strongly reflect those interests. The tech giant aims to blend its powerful algorithm with the user’s personal interests to deliver more appealing results.
The tool brings a new direction for YouTube discovery
The feature is still not widely accessible. YouTube has also not yet offered clear details about how the prompts are processed or how heavily they influence the underlying algorithm. However, the tool likely does not alter the actual usage-based recommendations. It just wants to work as a companion to deliver more precise suggestions. The tool is still kept under the experimental features, and PCMag reports it was unavailable across several tested accounts .
If you are included in the test, accessing the tool is pretty simple. All you have to do is click on the “Your Custom Feed” option and then describe the kinds of videos you want to see more often in the prompt box. The system will now incorporate the prompt into future recommendations. Also, do not get confused this tool with the Customize your feed feature introduced in 2024. The latest tool uses AI to deliver suggestions, whereas the older one relies on upvoting and downvoting sample videos.
