Social Feeds Are Becoming User-Directed Discovery Engines

Meta introduced new AI-powered controls for Facebook Feed and Reels on 13 October, allowing people to tell its recommendation systems what they want to see in their own words. The Your Algorithm feature lets users add or manage topics, while follow-up prompts after “Interested” or “Not interested” feedback can refine recommendations in real time.

The update changes more than the settings menu. Facebook has long inferred preferences from behaviour such as likes, comments and watch time. It is now adding freeform language to the ranking system, giving audiences a direct role in defining relevance.

For CMOs, broad engagement could become a less complete guide to distribution. Content will still need to perform, but it may also need a clear topic and audience promise that users can actively request.

What Meta Changed

Meta said Your Algorithm will be available through the three-dot menu on Facebook posts. Users can enter topics they want more or less of, and the system is designed to respond within the same session across Feed and Reels.

The company said its ranking systems draw on “thousands of other signals” in addition to actions such as likes, comments and viewing. Large language models now let Facebook combine those behavioural signals with interests, context and intent expressed in natural language.

Meta described the release as putting people “in the driver’s seat.” The platform is not abandoning prediction. It is supplementing prediction with instructions.

AI Commoditizes Content: Why Brands Must Compete on Taste, Not Volume

As generative AI floods markets with volume, smart brands are competing on taste and quality instead. Here's how leaders like Veo Sports and Unilever are winning.

Mission MediaMission Media

Why Inferred Signals Are No Longer Enough

Facebook is following a direction already visible elsewhere in Meta. Instagram expanded Your Algorithm controls to Feed in June, after introducing them for Reels, allowing users to inspect, add and remove topics associated with their interests. Instagram head Adam Mosseri framed the limitation clearly: “The system learns from what you tap, watch, and share, but you don’t really get to tell it what you want.”

The shift is arriving alongside pressure for stronger user choice. Australia proposed legislation in September that would require platforms to offer a choice between algorithmic recommendations and content from followed accounts. Prime Minister Anthony Albanese said platforms would have to provide “genuine and enduring choice” over feeds.

An August YouGov poll of 1,116 US adults found Facebook users preferred chronological feeds to algorithmic ones by 49% to 26%. Among heavy Facebook users, however, algorithmic feeds narrowly led chronological feeds, 41% to 38%. The commercial challenge is not simply to remove algorithms, but to make them feel responsive and accountable.

How Brands Compete In A Prompted Feed

Explicit preference inputs could alter how social content earns discovery. A post that attracts incidental engagement may not build lasting distribution if audiences request less of its topic. Conversely, a clearly positioned series may benefit when users describe the subject they want and the system can match that request to recognisable content.

That puts greater weight on editorial clarity. Social teams need consistent language for the themes a brand owns, while creative needs to make those themes legible without becoming mechanical. The objective is not to stuff captions with keywords. It is to make the value of watching or requesting the content easy to understand.

This may sharpen the difference between temporary reach and durable preference. Viral distribution can expose a brand widely, but user-directed feeds reward a more demanding outcome: people choosing to see more of a subject associated with that brand.

What Regional Marketing Leaders Should Do Next

The APAC implication is operational. Users across the region may describe the same interest differently by market, language and cultural context. A single regional content taxonomy may be too blunt for a recommendation system that interprets natural-language intent.

Regional marketing leaders should review how campaign themes are named locally, then test whether those themes remain recognisable across posts, creator briefs and community management. Measurement should distinguish between one-off engagement and repeat exposure within a defined topic, even if platforms do not disclose every ranking input.

Meta’s release suggests social discovery is becoming partly user-programmed. The executive decision is to treat audience language as a distribution input, not merely copy feedback. Brands that can state clearly what they are useful for may be easier for both people and ranking systems to choose.