Meta introduced new controls for Facebook Feed and Reels on 13 October 2026, including a tool called Your algorithm that lets people name the topics they want to see. The company says people can also describe, in their own words, what they want more or less of after responding to a post, with recommendations changing during the same session.
Key points
- Your algorithm gives people a place to add and manage topics for Feed and Reels.
- Meta may ask for written detail after someone selects ‘Interested’ or ‘Not interested’ on a post.
- Meta says large language models help its ranking systems use those words alongside engagement signals.
Meta adds words to engagement signals
Facebook’s ranking systems already draw on actions that can indicate interest, including posts someone likes or comments on and what they watch. Meta says those actions are among thousands of signals its systems use when deciding what to recommend. With the new controls, someone can state a preference directly when they want their recommendations to change.
Meta says large language models help its ranking systems use the interests, context and intent in a person’s written instructions. In plain terms, the system can take account of a request phrased as a sentence, rather than requiring every preference to arrive as a tap on a predefined topic. Meta describes the change as combining freeform text with the information its systems already gather about what people enjoy.
Choosing what to watch after an interest changes could therefore involve stating that change instead of waiting for viewing behaviour to convey it. That would give the ranking system a different kind of input from the history of likes, comments and watched posts Meta describes. The company says its recommendations respond within the same session, so the written request is intended to affect what comes next, rather than serve only as a preference for a later visit.
The distinction matters because the two kinds of input answer different questions. A like or a stretch of viewing records an action already taken, while a written request can describe what someone wants now. Meta says it uses both in ranking Feed and Reels. Its account of the change is about how preferences enter those systems and how quickly recommendations respond, rather than about replacing the signals the systems already use.
Your algorithm gives topics a place
Meta is rolling out Your algorithm as a place to add and manage topics for Facebook Feed and Reels. The tool is available through the three-dot menu, according to the company. It offers a route for setting interests apart from reacting to an individual post: someone can name a subject they want to see, rather than relying on a series of interactions for the ranking system to infer it.
The topic tool and the written feedback serve related purposes, but start from different places. Your algorithm begins with subjects a person chooses to manage. The feedback options begin with a post already on screen and allow a more detailed response to it. Both send explicit direction into systems that Meta says have long used behaviour, such as likes and viewing, to work out interests.
For companies whose posts appear in Feed or Reels, this puts another form of audience preference into the recommendation process Meta describes. The company says the algorithms use freeform text together with existing information about what people enjoy, so a written request joins the inputs used to rank content. Meta’s announced controls concern recommendations within Facebook Feed and Reels, rather than the creation or editing of the posts themselves.
Facebook adds questions to post feedback
Meta is also changing the response available on posts in Feed and Reels. After a person selects ‘Interested’ or ‘Not interested’, the company may ask a follow-up question about what they liked or disliked. A person can then say what they want to see more or less of in their own words. Meta says recommendations will update in real time from that input.
That sequence gives the initial response and the explanation separate jobs. ‘Interested’ supplies a broad signal about the post; the follow-up offers room to describe the preference in more detail. Meta says its large language models can use context and intent in natural language, which is why a written explanation can become part of ranking rather than remaining only a comment attached to a piece of feedback.
Meta has updated the three-dot menu on every post to make the customisation controls easier to reach. Your algorithm can be found there, while the additional feedback starts from the ‘Interested’ and ‘Not interested’ choices on posts. Those entry points put the new controls alongside the content being recommended, where a person can choose a topic or respond to what has just appeared.
