Facebook lets users tell its algorithm what to show in their own words

Meta is rolling out new AI-powered tools that let Facebook users tell the platform, in their own words, what they want to see in Feed and Reels. According to Meta, recommendations respond right away to what users write.
- Meta introduced AI-powered tools that let Facebook users describe in their own words what they want to see in Feed and Reels.
- A new feature called Your algorithm lets users add and manage topics, and is found in the three-dot menu on posts.
- After clicking 'Interested' or 'Not interested,' Facebook may ask follow-up questions, and recommendations update in real time, according to Meta.
- Meta says large language models let the ranking systems use natural language alongside existing signals such as likes, comments and watch behavior.
A dedicated space called Your algorithm
At the center of the update is a feature called Your algorithm. Meta describes it as a place where users can directly add and manage the topics they want to see in Feed and Reels, or in the company's words, "a dedicated space to tell us exactly what you're into."
The feature can be found in the three-dot menu, which Meta says it has updated on every post to make the new customization tools easier to reach. Meta says the aim is for users to fine-tune their experience and interact directly with the algorithms that rank and recommend what they see on Facebook.
Feedback directly from posts
Meta is also adding more ways to give feedback on individual posts in Feed and Reels. After a user clicks 'Interested' or 'Not interested,' Facebook may ask follow-up questions to get more detail on what the user liked or disliked.
Users can then say, in their own words, what they want to see more or less of. According to Meta, recommendations will update in real time. The company calls this "a more direct, conversational way to shape your experience."
Language models in the ranking systems
Meta says its existing ranking systems already infer interests from what people do on Facebook, from the posts they like and comment on to what they watch, among thousands of other signals. The new element is the use of large language models (LLMs).
According to Meta, LLMs can incorporate the interests, context and intent behind natural language, which makes possible "an algorithm you can actually talk to." Users can state what they want more or less of, and recommendations adapt within the same session to reflect what matters to them at that moment.
Meta describes this as a real evolution in how ranking works. The system brings freeform text together with what Facebook already understands about a person's tastes, and the company says the algorithms behind Feed and Reels should keep becoming more relevant and responsive over time.
What it means for users
For Facebook users, the change shifts some control from purely behavior-based inference toward explicit instructions. Instead of relying only on likes, comments and watch history, people can now state their preferences in plain language and see the results within the same session, according to Meta.
Meta's announcement does not give a rollout schedule or list the markets where the tools are available, saying only that it is "rolling out" the features.


