Facebook lets you tell the algorithm in your own words what you want to see

Facebook users will soon be able to talk directly to the algorithm that decides what they see. On October 13, 2026, Meta announced a tool called "Your algorithm," which lets people describe in their own words what they want to see more or less of in Feed and Reels.

Meta itself describes the feature as an algorithm "you can actually talk to." Here, an algorithm is the computer program that chooses which posts and videos are shown to you.

How it works

"Your algorithm" is a dedicated space where users can add and manage the topics they want to see in Feed and Reels. It is accessed through the three-dot menu on posts.

When users tap "Interested" or "Not interested" on a post, Facebook may ask follow-up questions. There, users can elaborate in their own words on what they would like more or less of.

Meta says large language models (the type of AI behind chatbots) are used to understand interests, context and intent in the text. This works alongside the signals Facebook already uses, such as likes, comments and viewing history. According to Meta, recommendations are updated in real time, meaning within the same session.

Also on Instagram and Threads

According to TechCrunch, Instagram got a similar feature in June 2026, while the version for Reels arrived in December 2025. Threads has a comparable tool called "Dear Algo."

Meta says AI language models in its recommendation systems in 2026 are linked to a 10 percent increase in viewing time for Instagram Reels. The figure comes from Meta itself and has not been independently verified.

Open questions

Meta's announcement does not clearly specify which countries the feature is coming to, or when. The news outlet MediaNama found in testing that the feature was not available in India on any Meta platform. It is therefore unclear when Danish users can expect it.

MediaNama has also asked Meta whether users' text is stored, whether it is used to target ads, and whether the feedback is used to train AI models. No answer had been received yet.

According to MediaNama, an earlier test of Threads' "Dear Algo" showed unexpected content in unrelated languages. That raises doubts about how consistently these kinds of tools work.