Facebook lets you talk directly to its recommendation algorithm

For years, training your social media feed has felt like trying to train a stubborn puppy. Linger on one weird video for five seconds by accident, and your feed gets flooded with identical clips for the next three weeks.

Now, Meta wants to let you set the record straight using plain English.

Meta announced the rollout of a new AI-powered feature called Your Algorithm for Facebook Feed and Reels. Instead of relying purely on silent behavioral clues, you can now literally type out what you want to see more or less of.

Talking back to the feed

The new tool lives inside the three-dot menu on any post. From there, you can access the Your Algorithm hub to add, remove, and manage topics you care about.

Facebook is also making feedback conversational. When you tap Interested or Not Interested on a post, the platform might follow up and ask why. You can respond with freeform text, like asking for more home workout tips but fewer supplement ads, and the recommendation engine updates in real time within the same browsing session.

Under the hood, Meta is leaning on large language models to parse the context and intent behind your natural language prompts, combining that input with the thousands of implicit signals it already tracks.

As reported by TechCrunch, this is part of a broader platform rollout. Meta first tested the feature on Instagram Reels back in December 2025 before expanding it across Feed and Explore in June. Threads rolled out a similar tool called Dear Algo, and now Facebook is bringing the same capability to its massive user base.

Why it matters

For everyday users, this cuts down the endless frustration of dealing with a recommendation black box. If your interests shift, or if the algorithm locks onto something you hate, you no longer have to spend days aggressively skipping posts to fix your feed. You can just tell it to stop.

Across the industry, Meta is nudging social media recommendation away from purely passive, dopamine-driven consumption toward something closer to interactive curation. If this works, rivals like TikTok and YouTube will face growing pressure to offer explicit, conversational controls instead of treating their ranking engines as untouchable secret sauce.

We think using LLMs for feed management is a smart design choice. It spares people from navigating endless setting menus and rigid category checklists. A single sentence can convey a level of nuance that traditional toggle switches never could.

The real question is how Facebook handles the inevitable clash between what people say they want and what they actually watch. Most people claim they want educational documentaries, yet their watch time says cat memes and street fights. If Facebook prioritizes watch time over typed preferences when the two inevitably disagree, the feature could easily end up feeling like a placebo.