For about fifteen years, the main way to influence a social feed was to behave in ways the feed could interpret, then hope. Linger on a video, regret it, and watch the platform conclude you wanted a hundred more. Meta now says Facebook users can just tell the algorithm what they want, which is a neat reversal of the usual arrangement.
The feature is called Your Algorithm. Meta announced it on Tuesday, and it covers both Facebook's Feed and Facebook Reels. It arrives alongside a broader claim that large language models are taking a bigger part in how Facebook ranks content.
What Your Algorithm actually does
The pitch is simple. Instead of relying only on automated guesses about your tastes, you can add and manage topics you want to see, and mark others as things you don't. Meta says recommendations "respond right away" once you make a change.
The controls come in a few forms:
- A dedicated tool. Your Algorithm is its own feature, where you can manage the topics you want more or less of.
- The three-dot menu. On any individual post, the menu now offers access to Your Algorithm, plus "Interested" and "Not interested" options.
- Buttons under posts. The "interested" and "not interested" prompts may also appear beneath posts, rather than being buried in a menu.
- Follow-up questions. Meta says Facebook may ask for more detail on what you liked or disliked, and you can answer in your own words.
That last item matters most. Tapping "not interested" is a blunt instrument, since it doesn't say whether you disliked the topic, the poster, or just the mood of the afternoon. A free-text answer at least gives the system something to work with.
The LLM part
Meta frames Your Algorithm as one piece of a set of AI-powered tools for personalizing Facebook. It also says it is using LLMs more in its ranking systems.
The company's reasoning is that language models can take in the interests, context, and intent behind natural language. In Meta's words, that unlocks "an algorithm you can actually talk to." You say what you want more or less of, and recommendations are supposed to adapt within the same session, reflecting what matters to you right now rather than what mattered last Thursday.
Meta has not, in the material available here, said how heavily the models are weighted in ranking, or how a typed preference is balanced against everything else the system already knows about you. The promise is responsiveness. The mechanics are left to the reader's imagination.
Not Facebook's first run at this
Facebook is arriving late to a feature Meta has already tried elsewhere. Instagram introduced Your Algorithm in June, extending customization across Feed and Explore. That followed an earlier launch limited to Reels in December 2025. Threads has its own related feature, called Dear Algo, which also lets users influence what gets recommended.
So this is less an experiment than a rollout. Meta tested the idea on Reels, widened it on Instagram, gave Threads a variant with a friendlier name, and has now carried it to its oldest and largest social product.
A nod toward streaming
There is a broader read on the move. As AI plays a larger part in how people find content, the traditional recommendation model, where the platform infers everything and the user submits, is looking dated. Giving people direct input moves social feeds closer to streaming services, where you can tell the service what you like and expect it to act on that.
It is also a mild irony that the answer to an opaque algorithm is a text box addressed to the algorithm. The platform still decides what to show you, and you are now allowed to lobby it. Whether that feels like control or like a more polite form of being managed will depend on how well the changes actually stick.
What to watch
The claim worth testing is the speed. Meta says updates register in real time, within the same session. That is easy to say and easy to check. If you tell Facebook you want fewer posts about a topic and they keep arriving, the feature has a credibility problem that no amount of natural-language branding will fix.
The other open question is how durable a preference is. A real-time adjustment is useful if it holds, and a gimmick if it fades by tomorrow. Meta describes adaptation to what matters "right now," which hints at a system tuned for the moment. Users who want a lasting change in their feed may find that the interface promises more immediacy than permanence.
For now, the practical advice is modest. Open the three-dot menu on a few posts, use the new buttons, answer the follow-up questions if they appear, and see whether Feed and Reels notice. After years of feeds that guessed in silence, being asked is at least a change.