Meta adds natural language algorithm tool to Facebook
Meta has introduced "Your Algorithm" to Facebook, allowing users to customize Feed and Reels using natural language prompts to directly steer automated recommendations.

Meta launched an AI-driven customization feature called Your Algorithm on Facebook on Tuesday, granting users direct control over their content recommendations through natural language prompts. Rather than relying exclusively on passive tracking metrics such as watch time, likes, or comments, the platform now allows individuals to write out specific instructions describing the topics and media types they want to see more or less frequently across their main Feed and Reels. Adjustments made through the system take effect in real time.
To support this conversational feedback loop, Meta updated Facebook's standard three-dot menu located on individual posts. Opening this menu now provides quick access to Your Algorithm, along with simplified Interested and Not interested options. When users select these controls, Facebook may issue follow-up questions to gather more detailed qualitative input, giving individuals a structured way to clarify exactly what elements of a post resonated or fell short.
The update expands Meta's multi-platform strategy to re-architect how recommendation engines discover and distribute media. Instagram originally debuted Your Algorithm for its short-form video surface, Reels, in December 2025, before expanding the natural language feature across its Feed and Explore sections in June. Additionally, Meta introduced a related prompt-based feed customization tool on Threads known as Dear Algo.
For platform developers, social marketers, and recommendation engineers, Meta's pivot represents an important transition from implicit behavioral modeling to explicit zero-party feedback. By mimicking personalization mechanisms common in digital streaming services, social media discovery is becoming actively user-steered. Content strategists must recognize that distribution will increasingly depend on explicit subscriber interests rather than passive algorithmic inference alone.
This is our own summary of reporting by TechCrunch AI



