Facebook adds freeform instructions for Feed and Reels recommendations

Tell Facebook which topics matter

Meta's Facebook announcement introduces AI-powered tools for shaping Feed and Reels recommendations in the user's own words. The company says its updated three-dot menu makes the controls easier to find on posts.

Your algorithm provides a dedicated place to add and manage topics a person wants to see. The feature is described as rolling out, so its appearance may depend on the account's current access.

Feedback can include an explanation

After selecting Interested or Not interested, a user may receive follow-up questions about what they liked or disliked. Meta says freeform feedback can tell the system what to show more or less of, with recommendations updating in real time.

The company describes large language models interpreting interests, context and intent from natural language alongside its existing engagement signals. That combines explicit instructions with information inferred from likes, comments and viewing, rather than replacing all other ranking inputs.

These are changes to recommendation controls. The announcement does not provide a measured result for every user or describe them as a chronological-feed setting.

Give the system a specific distinction to learn

Franklin analysis: a useful trial should separate the topic from the kind of treatment desired. Someone might want more practical gardening advice while wanting fewer repetitive promotional posts about gardening. A freeform explanation can express that distinction more clearly than a single positive or negative click.

After changing a topic or supplying feedback, inspect several recommendations and check whether the adjustment matches the request. Revisit saved topics when interests change. The practical benefit is a more direct way to express preferences, but the outcome still depends on how the ranking system interprets those preferences alongside its other signals.