Meta has introduced a new feature called Your Algorithm for Facebook, which leverages artificial intelligence and large language models. This update allows users to provide direct input in natural language to customize their content recommendations for both the main feed and Reels.
Previously, Facebook relied on behavioral signals such as likes, comments, shares, and watch time to infer user interests. The new system shifts this approach by enabling users to explicitly describe their preferences, allowing the algorithm to adjust recommendations in real time based on these specific instructions.
The company has updated the three-dot menu next to content posts, providing a new interface for personalized settings. Beyond standard interest toggles, the system now asks for specific feedback on why a user likes or dislikes certain content, allowing for more granular control over the feed.
Meta states that large language models are now integral to its recommendation architecture. These models enable the system to interpret user intent more accurately and explain the logic behind content selection, moving away from purely behavior-based predictions toward a more interactive model.
