Loading page…
Meta says recommendations adjust within the same session. How much influence written requests have over the final mix remains unspecified.
Listen to this story
Meta is extending user-directed recommendation controls to Facebook, giving people a way to state interests instead of relying solely on activity-based inference. The October 13, 2026 rollout adds written preferences and post-level feedback for Feed and Reels; Meta says language models translate requests into ranking signals. Behavior signals remain part of the mix, however, and Meta has not disclosed how much weight explicit instructions receive, so the practical degree of user control is still unclear.
Users can manage topics through Your Algorithm in a post’s updated three-dot menu, or mark posts Interested or Not interested and explain their reactions.
Meta first introduced Your Algorithm on Instagram Reels, later extending controls to Feed and Explore; Threads has a related feature called Dear Algo.
Meta says recommendations can adapt within the same session, but that response speed has not been independently measured.
Facebook users can now tell the platform what they want to see, rather than leave it to infer their interests from their activity. Meta began rolling out Your Algorithm on October 13, 2026, letting people shape Feed and Reels recommendations with written preferences and feedback on individual posts.
Facebook’s existing recommendation systems draw on what people like, comment on and watch, among other signals. Those actions help the platform infer interests without requiring someone to describe them. Written instructions do not replace Facebook’s other inputs. Meta says it combines freeform text with what the platform already understands about someone’s interests.
Threads also has a related recommendation-control feature called Dear Algo. Facebook’s launch brings this approach to another Meta platform.
The entry point is Facebook’s updated three-dot menu on posts. From there, users can open Your Algorithm, a dedicated space to add and manage the topics they want recommended. This lets people specify interests directly, without waiting for a relevant post to appear.
The same menu offers Interested and Not interested feedback on individual posts. Meta says Facebook may then ask follow-up questions about what someone liked or disliked. Users can answer in their own words about what they want more or less of, explaining the preference behind a positive or negative reaction.
Behind the interface, Meta says large language models—AI systems that process language—help its ranking systems interpret the interests, context and intent in a written request. The aim is to make a person’s explanation usable by the systems that rank and recommend posts, rather than simply collect topic names.
Meta says recommendations respond in real time, adapting within the same session. That is the company’s promised response speed, not an independently measured result. Its announcement does not specify how much weight a written instruction receives compared with existing behavior signals, leaving open how strongly users’ words govern the final mix.
Loading discussion...
Join the conversation
Explain which signal better reflects what you want in your feed.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.