Tired of AI slop? Facebook gives users more control over Feed and Reels
Facebook is introducing artificial intelligence (AI)-powered tools that allow users to customise the content recommended to them in Feed and Reels sections on the platform. According to Facebook, the new features let users tell the platform what they want to see more or less of, with recommendations adjusting in response. Users can also manage their preferred topics through a new section called “Your algorithm” and provide more detailed feedback on individual posts.
The move comes as social media platforms explore ways to give users greater control over recommendation systems, with Threads having introduced a similar feature called “Dear Algo”.
Users can tell Facebook what they want to see
The new “Your algorithm” feature gives users a dedicated space to add and manage topics they want to see in their Facebook Feed and Reels. It can be accessed through the three-dot menu on posts.
For instance, users who want more technology news, travel videos or cooking content can communicate these preferences directly rather than relying entirely on their viewing and engagement history to influence recommendations.
Facebook is also expanding the feedback options available on individual posts. Users can select “Interested” or “Not interested” and may be asked follow-up questions to explain their choices.
They can then describe, in their own words, what they want to see more or less of. Meta said recommendations would respond in real time, allowing users to refine their experience without waiting for the platform to infer their preferences from their activity. The updated three-dot menu is intended to make these customisation options easier to access.
Large language models power the new tools
Facebook has traditionally relied on signals such as likes, comments, shares and viewing behaviour to determine which posts and videos appear in users' feeds. Its ranking systems analyse thousands of signals to estimate what individual users may find relevant.
The new tools bring large language models (LLMs) into this process, allowing the system to interpret instructions expressed in natural language.
Unlike previous feedback mechanisms, which largely depend on users interacting with content, the new approach allows them to explain the context behind their preferences. For example, a user could ask for more smartphone reviews but fewer promotional videos, helping the system distinguish between related types of content.
According to Meta, the recommendations can adapt within the same session as users provide feedback. The company said combining text-based instructions with existing signals could make recommendations more relevant and responsive.
However, the announcement does not mean users can completely determine everything that appears in their feeds. The new controls allow them to influence recommendations, while Facebook's ranking systems continue to determine which content is shown.
Threads has a similar feature
Meta's announcement follows the introduction of “Dear Algo” on Threads, which allows users to influence their recommendations by telling the platform what content they want to see more or less of.
Both features reflect a shift towards more direct user input in recommendation systems. Instead of relying only on inferred preferences, platforms are allowing users to communicate their interests explicitly. Their effectiveness will depend on how accurately the systems interpret instructions and how consistently those preferences shape the content users see.