Facebook enables feed personalization through AI: new features

Facebook introduces new AI tools: control your recommendations feed with text

Facebook (Meta) has begun rolling out innovative artificial intelligence-based tools that will give users unprecedented control over how their news feeds and Reels are shaped. From now on, users will be able to directly influence the content they see using a simple text description of their preferences.

“Your algorithm”: your personal feed manager

The main innovation is the introduction of a separate section called “Your algorithm.” In this section, users will be able to create, add, and edit a list of topics they want to see in their recommendations. This will make it possible to fine-tune the feed much more precisely to individual interests.

Access to the “Your algorithm” section will be available through the three-dot menu located next to each post. The company has also updated this menu to make navigation easier and provide quick access to recommendation settings.

The setup process includes the option to add or remove specific areas of interest. An important new feature is a text field that appears when selecting the “Interested” or “Not interested” options. For example, a user will be able not only to mark celebrity news as uninteresting, but also to specify the preference by asking to see more tech reviews instead. The algorithm will take these instructions into account in real time, instantly rearranging the order of posts.

From mathematical models to understanding human language

Previously, Facebook’s ranking systems were based mainly on complex mathematical models that analyzed thousands of behavioral signals to predict user engagement. This made it possible to anticipate what kinds of content might be interesting, but it did not always take into account the nuances of personal preferences.

The new AI tools are based on large language models. These models are capable of reading not only behavioral patterns, but also analyzing context, mood, and the user’s exact intent expressed in free-form text. Meta believes that this combination of behavioral analysis and direct text preferences will make the recommendation feed more flexible, personalized, and less intrusive.

This rollout is part of Meta’s broader efforts in artificial intelligence. Earlier, the company introduced two new AI models — Muse Wonder and Muse Wonder Full — designed to solve complex tasks in personal and professional settings, including programming and workflow automation.