Meta adds natural language controls to Facebook to refine feed recommendations

Article Summary

Meta has launched the AI-powered “Your Algorithm” feature on Facebook, enabling users to influence Feed and Reels content by describing their preferences in natural language. This tool, which provides real-time adjustments to post rankings, is complemented by an updated three-dot menu that allows users to express interest or provide detailed follow-up feedback on specific posts. The initiative mirrors similar personalization tools previously deployed on Instagram and Threads to offer users greater control over AI-driven content discovery.

AI Analysis

## Core Assessment

Meta is bringing its "Your Algorithm" feature to Facebook, allowing users to type natural-language instructions to reshape what appears in their Feed and Reels. Paired with an expanded three-dot menu that adds follow-up questions to standard "interested" or "not interested" taps, the platform promises real-time ranking adjustments based on explicit user input rather than purely passive engagement signals. This extends a pattern Meta already established on Instagram and Threads with its conversational "Dear Algo" experiment.

From an engineering perspective, letting users speak directly to a recommendation model sounds empowering, but it shifts an unpaid maintenance burden onto the consumer. Black-box recommendation engines struggle precisely because human tastes are nuanced and situational; expecting people to actively prompt-engineer their feed via follow-up surveys is an admission that passive behavioural tracking frequently degrades into engagement bait. While real-time weight adjustments give users a sense of agency, natural language is inherently ambiguous. Without seeing the underlying weights, users have no visibility into how long these explicit prompt overrides survive against Meta's primary algorithmic objective: keeping eyes on the screen to serve ads.

## Impact and Advice

* **Social Media Managers & Content Creators**: Audit distribution metrics over the coming weeks to identify shifts in organic reach across Feed and Reels. Do not overhaul production based on the assumption that niche audiences will successfully prompt their way to your content; maintain clear topical tagging so semantic classifiers can still match your posts when users issue broad natural-language requests.
* **End Users**: Use the conversational feedback controls selectively for persistent feed annoyances rather than attempting to curate an ideal feed. Explicitly prompt the system to suppress low-value topics via the updated three-dot menu, but recognize that conversational tuning requires ongoing correction whenever the recommendation loop drifts back toward generic trending content.

## Bottom Line

Meta is asking users to debug its recommendation system with natural-language prompts, but giving you the steering wheel does not mean they have changed the engine's destination.

Explore this story further