YouTube is developing an experimental feature called “Your Custom Feed” to address issues with its cluttered homepage feed. The platform aims to reduce user frustration from algorithm-driven recommendations that often fail to align with individual preferences, allowing users to input custom prompts directly on the app’s homepage.
The core challenge stems from YouTube’s recommendation algorithm, which has long struggled to accurately interpret user behavior. Reports indicate that the system frequently misinterprets casual viewing as deep interest. For example, if a user watches several Disney videos, the algorithm assumes a strong affinity for Disney content and proceeds to fill the feed with an excessive volume of similar videos, even when the user seeks variety or different topics. This overgeneralization leads to a homepage overwhelmed by irrelevant suggestions, prompting users to spend more time scrolling past unwanted content rather than engaging with desired videos.
In the ongoing experiment, participants see the “Your Custom Feed” option positioned adjacent to the familiar “Home” button on the YouTube homepage. Selecting this button opens an interface where users can enter specific text prompts based on their current interests. This active input mechanism shifts control from passive algorithmic curation to user-directed personalization. Rather than relying solely on past watch history, the feature uses these prompts to guide immediate and subsequent recommendations, creating a more responsive feed tailored to explicit requests.
To illustrate its functionality, a user interested in culinary content can type “cooking” into the prompt field. In response, YouTube elevates videos related to cooking topics, such as recipes, techniques, or tutorials, in the displayed feed. This process repeats with new prompts, enabling ongoing adjustments without resetting the entire viewing history. Such customization offers a direct alternative to traditional methods of refining the feed, like individually selecting videos and applying feedback options such as “Not interested” or “Don’t recommend channel.” These legacy tools require repeated manual interventions across multiple items, whereas prompt-based input streamlines the process into a single, targeted action.
Beyond YouTube, similar innovations appear on other social platforms. Threads has been observed testing a feature for configuring its algorithm settings, giving users options to tweak recommendation parameters. Likewise, X is developing a tool that permits users to tag its AI chatbot, Grok, in interactions, which in turn modifies the personalized feed based on those engagements.
