Facebook's Your Algorithm Lets You Directly Shape Feed Topics

Meta has brought Your Algorithm to Facebook, letting users add and manage the topics that appear in Feed and Facebook Reels. The launch was reported on Oct. 13, 2026, and extends the same control Meta launched on Instagram earlier this year, according to The Verge.

Access lives where ranking choices already appear. Users can open Your Algorithm from the three-dots menu on a post, or select "interested" or "not interested" from that menu. In some cases Meta also shows those "interested" and "not interested" buttons inline under posts.

Meta says Facebook recommendations "respond right away" after a topic is marked "interested" or "not interested" in Your Algorithm. It does not stop at the toggle. Your Algorithm may ask follow-up questions about what was liked or disliked, and users can describe in plain language what they want more or less of. Recommendations then update in real time.

Meta describes the feature as part of its AI-powered tools for shaping Facebook content, and says it is using LLMs, models trained to read and interpret text, more in ranking, the software that orders posts. Threads has a parallel version called "Dear Algo." The Transparency Center User Control page describes both in the same terms, as a way to shape what users see by entering topic preferences directly, according to the Transparency Center.

The broader context here will be familiar if you follow recommender systems. Most Feed ranking still runs on implicit signals, such as dwell time, skips, likes and replays. Explicit topic controls reverse that default. They trade a little user effort for clearer guidance the ranker can use when behavior is noisy or contradictory.

The system design question here is the natural-language loop. A binary toggle is easy to store and enforce. Free-text requests are harder. They must be parsed, mapped to topic lists, and managed when instructions conflict over time. Tying this to heavier LLM use in ranking suggests the text is likely read during ranking or reranking, the final ordering pass, where time budgets are tight and consistent interpretation matters.

In my view, the change is useful because it is modest. It does not replace behavior-based personalization. It gives experienced users a steering wheel for cold starts, topic shifts and cleanup after drift. I have watched my own children retrain their feeds this way when interests change, not through settings pages but by hitting every not-interested option until the system adjusts. A topic manager with instant feedback shortens that loop.

Worth flagging for builders is the measurement tradeoff. Direct controls improve perceived agency, which users say they like. They do not always lift short-term session numbers, especially if users narrow their own mix. The long arc favors controls that are easy to understand. Systems that can be corrected tend to keep trust longer than opaque systems that guess right more often.