On October 13, Meta announced a new feature called "Your Algorithm" for the Facebook Feed and Reels. The AI-powered tool lets users shape recommendations in their own words: they describe in text what they want to see in the feed, and the system adjusts recommendations in real time. It is the first major step toward letting Facebook users configure the recommendation mechanism directly.
The new capabilities open through an updated three-dot menu on every post. The menu includes a "Your Algorithm" section — a dedicated space for adding and managing the topics users want to see in the Feed and Reels. From this section, users add topics they care about directly and remove ones they don't.
Every post also gained "Interested" and "Not interested" buttons. After a user taps one of them, Facebook may ask follow-up questions to learn the details of what was liked or disliked. Meta calls this mechanism conversational feedback: the recommendation system no longer runs one way, but is tuned through a question-and-answer exchange with the user.
"Tell recommendations what you want to see more or less of — they'll respond in real time and adapt within the same session to what matters to you right now." — Meta's official announcement
Topic Management and Conversational Feedback
In the "Your Algorithm" section, users directly add and manage the topics they care about. Where feed recommendations used to be shaped only by indirect signals — likes, comments, watch time — users now express their preferences in open text. After tapping "Interested", the system asks additional questions to clarify what exactly was liked; with "Not interested", it asks what to show less of.
Meta emphasizes that recommendations update in real time: as soon as a user states a preference, the system adapts. That means recommendations can be dynamically reconfigured during a session — the feed changes the moment the user's interest changes. More in Meta's official announcement.
According to TechCrunch, the move is Facebook's latest step toward giving users more control over the algorithm than before. Where recommendations previously ran mostly on passive observation — the system guessing what was liked — the user now becomes an active participant: they say directly what they want to see.
The Technology: Large Language Models
Meta's official announcement notes that ranking systems previously inferred user interests from thousands of signals such as likes, comments, and viewed content. Large language models (LLMs) have now joined the process: they take into account interests, context, and intent expressed in natural language — so users can "talk" to recommendations in plain words.
The company calls the change a major evolution in how ranking works: a free-text preference merges with previously collected interest signals, so the algorithms behind the Feed and Reels grow more adaptive over time. Because large language models grasp subtle nuances of human language, recommendations are tuned more finely than in previous-generation systems.
The new feature is built on top of Facebook's existing ranking infrastructure: when a user enters text, the system combines it with signals from past behavior. That way, recommendations draw on both personal history and current preferences. According to Meta, the approach makes recommendations more personal, nuanced, and tailored to the individual user — the system no longer merely guesses what to show, but follows the user's direct instructions.
Experience on Instagram and Threads
"Your Algorithm" was first tested on Instagram: it launched in Reels in December 2025 and in the Feed and Explore sections in June 2026. Threads has a similar feature called "Dear Algo". So Meta first tested the idea of giving users direct control over the algorithm on other platforms and is now bringing it to the main Facebook Feed and Reels. According to TechCrunch, the Facebook version of the feature builds on that same experience — the concept is identical across all platforms: users say in their own words what they want to see, and the algorithm adapts recommendations in real time.
The new tools have been announced for the Facebook Feed and Reels. According to Meta, the system for controlling recommendations in your own words is the next generation of ranking models. The experience across all three platforms rests on one idea: the recommendation system listens to openly expressed user preferences. And the Facebook version targets the widest audience — the Feed and Reels are the social network's central surfaces.
