Facebook is giving users more control over their algorithm

Meta announced on Tuesday that it has brought "Your Algorithm," its AI-powered personalization control, to Facebook. The feature lets users steer what appears in their Feed and Reels by describing what they want to see more or less of in their own words, rather than relying solely on automated recommendations. Meta first introduced Your Algorithm on Instagram, and the Facebook rollout extends that same natural-language approach to its largest social app.
Mechanically, the tool is a conversational interface layered over the ranking systems that decide which posts and videos surface. Instead of only reacting to implicit signals such as watch time or taps, users can type free-form statements about their interests, and Meta says changes are reflected in their recommendations in real time. The company positions the feature as a way to make recommendations more relevant by letting people interact directly with the algorithms that rank and recommend content.
Meta has also reworked Facebook's three-dot menu on individual posts. From that menu, users can open Your Algorithm or choose "Interested" and "Not interested" on a given piece of content. According to Meta, the platform may ask follow-up questions to clarify what a user liked or disliked, so the feedback can be more detailed than a single toggle. That combination matters: the three-dot menu is a per-post, in-context action, while Your Algorithm is a broader, standing statement of preference. Together they give Facebook two feedback channels, one immediate and granular, one persistent and descriptive.
The launch fits a pattern across Meta's apps. Instagram introduced Your Algorithm in June, extending customization to Feed and Explore after first launching it for Reels in December 2025. Threads has a related feature called "Dear Algo" that similarly lets users influence recommended content. In other words, Meta is standardizing a single interaction model, natural-language preference statements, across several products rather than treating each app's recommendation controls as separate systems.
For developers and product teams, the notable design choice is the use of natural language as the control surface for a recommender. Conventional personalization controls are usually structured: categories, sliders, topic lists, mute keywords or simple thumbs up and down. A text box is fuzzier to parse, but it lets users express intent that taxonomies do not capture, such as wanting less of a broad theme rather than a specific account. Making changes appear in real time implies the described preferences feed into the ranking pipeline without a lengthy retraining cycle, though Meta has not detailed the underlying implementation. This suggests the preferences likely act as signals or adjustments applied alongside existing ranking models rather than as full model retraining.
This is also a shift in how the product is framed. Recommendation systems have historically been opaque, optimized for engagement and tuned by internal teams. Bringing users into the loop through conversation moves Facebook closer to the model familiar from streaming services, where viewers can tell a platform what they like to refine suggestions. Whether users treat the feature as a genuine lever or as a novelty depends on how faithfully the stated preferences translate into real changes in the feed. A likely trade-off is tension between what users say they want and what keeps them engaged; the source does not say how that is resolved, but it is the central question any such control raises.
There are open practical questions the announcement does not answer. Meta has not said how strongly a stated preference weighs against behavioral signals, whether preferences persist across sessions and devices, or how they interact with advertising. It has not described availability beyond the announcement, nor whether the feature is rolling out broadly or gradually. For now, the concrete details are the feature's name, its presence in Feed and Reels, the natural-language input, real-time reflection of changes, the updated three-dot menu with Interested and Not interested options, potential follow-up questions, and the prior Instagram and Threads precedents.
Why it matters: Meta is making natural-language preference control a shared interaction pattern across Facebook, Instagram and Threads, which affects anyone building recommendation, ranking or personalization interfaces. If users can describe intent in words and see results quickly, product teams may face rising expectations for similar controls, and developers should watch how Meta handles feedback that conflicts with observed behavior, since that is where a stated-preference system either earns trust or quietly fails.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Meta announced the launch of its new AI-powered feature, “Your Algorithm,” a feature it first launched on Instagram.