Facebook is rolling out a new control called Your Algorithm that lets people describe, in ordinary language, what they want to see more or less of in Feed and Reels. Meta announced the feature on October 13, placing it in the three-dot menu attached to posts and promising that recommendation changes will take effect within the same session.
The feature gives Facebook users a more direct way to correct the platform’s guesses. Instead of hoping that a few likes, skips or hidden posts eventually reshape a feed, a person can add and manage topics or explain why a particular recommendation is unwanted. Meta is also adding follow-up questions after the existing Interested and Not interested actions.
That is meaningful control, but it is not a switch that turns Facebook into a chronological, friends-only service. Your Algorithm adjusts recommendations inside Meta’s ranking system; it does not replace that system, erase the other signals Facebook has collected or change every part of the service.
How to find Facebook’s Your Algorithm control
Meta says the control is available from the three-dot menu on posts in Feed and Reels. Because the company describes the feature as rolling out, it may not appear for every account immediately.
- Open Facebook and find a post or Reel related to a topic you want to tune.
- Select the three-dot menu on that item.
- Look for Your Algorithm to open the topic-management screen.
- Add or manage topics you want Facebook to recommend.
- For a specific item, choose Interested or Not interested. If Facebook asks for more detail, describe what you want more or less of in plain language.
- Check the next set of recommendations. Meta says the ranking system can respond in real time, including during the same session.
The wording matters. A broad instruction such as “less sports” gives the system less to work with than “show me fewer professional football highlights, but keep local running events.” Meta’s stated reason for using a large language model is that it can interpret topic, context and intent from freeform text rather than treating every preference as a simple yes-or-no label.
What changes under the hood
Facebook has long inferred interests from likes, comments, viewing time, skips, follows and many other behavioral signals. It also offered direct controls such as Show More, Show Less, Hide and Snooze. Your Algorithm adds an explicit text signal to that mix.
According to Meta’s announcement, the new text is combined with what Facebook already understands about a person’s interests. It does not become the sole instruction that ranks the feed. That distinction helps explain why one request may change recommendations quickly without producing a perfectly clean reset.
The scale of the existing system is considerable. In a technical explanation of its recommendation pipeline, Meta described narrowing tens of billions of possible items to thousands and then a few hundred candidates before ranking the final selection. The system analyzes text, images, audio and video, and it draws on both explicit feedback and passive behavior such as watching a clip to completion or leaving after a few seconds.
Your Algorithm is designed to make a user’s stated intent another high-value input. It is especially useful when behavior is ambiguous. Watching an upsetting video to understand it, for example, can look like interest to an engagement model; a written instruction can tell Facebook that the subject should appear less often.
What Your Algorithm does not promise
Meta’s announcement applies to recommendations in Feed and Reels. It does not say that the feature will create a friends-only feed, switch every post to chronological order, eliminate sponsored posts or manage advertising preferences. Facebook’s separate feed and privacy settings still matter.
It also is not a data-deletion tool. Earlier in 2026, Meta expanded how information supplied by other businesses can be used to personalize non-ad content, including Feed and AI responses. The company directs users to a separate Activity from other businesses setting for that data use. Changing a topic in Your Algorithm should therefore be understood as a recommendation preference, not as a request to delete an activity record or prevent Meta from receiving information from another business.
Nor does the tool guarantee that every request will be followed literally. Ranking systems balance many inputs, including the accounts and groups a person follows, content eligibility rules, predictions about engagement and variety. Meta has not published a precise weighting for freeform instructions, a retention period for the text people enter or a complete list of supported languages and regions in the launch announcement.
How to get better results
The best use of the feature is likely to be a series of precise corrections rather than one sweeping command. Name the unwanted subtopic, format or context, then preserve the part you still value. “Less celebrity breakup speculation, more film reviews” is more actionable than “fix my feed.” For Reels, distinguishing between a subject and a format can also help: a user might want fewer reaction videos without removing the underlying topic.
Use the post-level controls alongside the topic manager. Marking an individual recommendation as Not interested supplies an example; the written follow-up explains what was wrong with it. If irrelevant posts keep returning, review followed Pages and Groups, use Unfollow or Snooze where appropriate, and revisit the broader Feed settings rather than relying on a single prompt.
People should also avoid entering sensitive personal information just to improve recommendations. A preference can usually be expressed as a content request without disclosing a diagnosis, financial problem or other private circumstance. “Show fewer weight-loss posts” communicates the ranking instruction without explaining why.
Why this change matters
Recommendation feeds increasingly ask users to train opaque systems through behavior. Facebook’s new control makes that relationship more explicit: people can state an interest directly, while the platform can translate that language into a ranking signal.
The approach has already spread across Meta’s products. Instagram expanded its own Your Algorithm controls beyond Reels to Feed and Explore, while Threads introduced a related feature called Dear Algo. Facebook brings the idea to a service whose Feed combines friends, groups, Pages, recommendations and advertising, making the limits of the control as important as its convenience.
If Meta’s real-time claim holds up in everyday use, Your Algorithm could shorten the frustrating cycle of hiding posts and waiting for the feed to adjust. The practical test is whether specific instructions keep working after the current session and whether users can understand, review and reverse the interests Facebook derives from them.
