Facebook has launched a new personalization feature called “Your Algorithm,” allowing users to dictate their content preferences using natural language rather than relying solely on passive behavioral tracking. The tool enables users to type specific descriptions of what they want to see more or less of within their Feed and Reels.
The update marks a transition from “inference-based” recommendations—where the system guesses interests based on clicks and watch time—to an intent-based model. By providing a text-based interface, Meta is giving users a direct way to reset or refine their recommendations without having to manually interact with dozens of posts to “train” the system.
How to Use Your Algorithm Topic Controls
The feature is integrated directly into the existing Facebook interface. To access the controls, users can tap the three-dot menu located on any individual post or Reel. From there, a new “Your Algorithm” option allows for the entry of specific topical feedback.
Because the system is powered by Large Language Models (LLMs), it can process the context and intent behind conversational phrases. For example, instead of just selecting a “Cooking” category, a user could specify “show me more vegetarian recipes but fewer dessert videos.” According to Meta, these updates are designed to respond in real-time, often altering the content of a user’s Feed within the same browsing session.
Integration and Limitations
The rollout follows the implementation of a similar feature on Meta’s Threads platform, known as “Dear Algo.” “Your Algorithm” provides granular control over the topics that appear in organic Feed and Reels content, allowing users to describe in their own words what they want to see more or less of.
By moving toward natural language feedback, Meta Newsroom suggests the platform aims to reduce the “algorithmic drift” that often happens when a user accidentally clicks on a video or post that does not reflect their actual long-term interests. This shift allows the ranking system to prioritize what users say they want over what they simply happen to click on.
