Meta announced on Oct. 13 that Facebook users can now describe in their own words what they want more or less of in Feed and Reels, and that large language models will read the text. The tool sits in the three-dot menu that Meta has updated on every post, in what the company calls a dedicated space to tell it exactly what a person is into.
Stay ahead on AI: Join 600,000+ professionals who read The Deep View, the free daily newsletter that keeps you up to date with artificial intelligence.
Subscribe to The Deep View free →
We may earn a commission if you sign up.
The change adds a typed signal to a ranking system that has so far guessed interests from behavior. Facebook already infers what people like from engagement such as likes, comments and watch time, and the new input lets a person state a preference outright instead of waiting for the system to work it out.
Three-Dot Menu Controls for Feed and Reels
According to Meta’s announcement, people can add and manage the topics they want to see from the menu, and tapping “Interested” or “Not interested” on a post may trigger follow-up questions about what was liked or disliked. Users can then say, in more detail, what they want more or less of. Meta says recommendations update in real time and can adapt within the same session, which makes a stated preference an immediate lever rather than a slow-moving one.
The announcement does not list example prompts, the exact wording of the follow-up questions, rollout dates or regions, and it carries no named spokesperson. Reporting by TechCrunch likewise found no staged rollout schedule and no stated limits on the feature.
Language Models as Interpreters of Typed Preferences
The technical change is in how the typed input is read. Meta says large language models interpret the interests, context and intent behind natural language, which lets the system handle a sentence of preference rather than a tap on a fixed category. Meta frames the result as an algorithm a person can actually talk to, and describes the typed text as an added signal for its existing ranking systems, not a replacement for them.
That framing is consistent with how Meta documents Facebook ranking. Its ranking and content transparency page says the system gathers inventory, evaluates thousands of signals, uses more than 100 prediction models to estimate whether a person will comment, share, click or watch, and combines those predictions into a relevance score. A typed preference becomes one more input to that score, with the same models still doing the ordering. The page also says predictions of an angry reaction, hiding, snoozing, unsubscribing or reporting are used to reduce a post’s distribution, and that the models are weighted differently by person and content type, with watch time mattering more for videos and clicks more for link posts. Meta limits the detail it publishes about signals on purpose, to prevent abuse.
Instagram Ran the Same Idea First
Facebook is the second Meta app to get the feature. TechCrunch reports that Instagram launched an algorithm-tuning control for Reels in December 2025 and extended it to Feed and Explore in June 2026. Search Engine Land’s coverage of the launch said it arrived in the United States on Dec. 10, 2025, with a Reels icon that opens a list of topics Meta’s AI believes a user has recently engaged with, plus options to see more or less of each and to add new topics. That report is the source for the Instagram version’s original scope.
Tessa Lyons, Instagram’s vice president of product, said at that launch that the company is always trying to show people the best possible reels for them, and she acknowledged that Instagram does not always get it right and that people’s interests change. TechCrunch adds that Threads offers a related feature called Dear Algo, which lets users shape their recommendations, so the typed-preference idea now spans three of Meta’s apps.
Open Questions About Persistence and Measurement
Several things remain unstated. An account of the Facebook rollout from Let’s Data Science notes that the announcement does not say how long a stated preference persists or whether it expires, and that no independent measurements of how much the controls change recommendations have been provided. No performance results for the new ranking experience have been reported either.
Meta also offers older tools that work on a post-by-post basis. The ranking page describes Show more and Show less, which temporarily raise or lower the ranking of a post and similar posts, as well as Feed Preferences for prioritizing Favorites or snoozing and unfollowing people, Pages and Groups. The ranking page also lists a Reduce control that adjusts how strongly problematic or low-quality content is demoted, and a Feeds tab that shows posts in reverse chronological order with ads included. The typed approach differs from all of these by letting a single sentence cover many posts at once, in Feed and in Reels, without a tap on each item.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
- Two passengers died at an Ohio toll plaza, and the NTSB now wants the cash lanes shut
- Long use of a common prostate pill is tied to a higher chance of glaucoma
- Ford is recalling 223,472 F-150 pickups because the fuel tank can leak or detach
- Hybrids have 15% fewer problems than gas cars, while EVs and plug-in hybrids have about 80% more, Consumer Reports finds
