Facebook Adds AI Controls That Let Users Tune Their Feed in Real Time

Meta is rolling out a new Facebook feature that lets people directly shape what shows up in their feed and Reels using AI-assisted controls. The update, called Your Algorithm, brings a more hands-on version of recommendation tuning to Facebook and is designed to make changes take effect immediately, which matters because it gives users a faster way to influence one of the platform’s most important discovery systems.

The company says the feature will let people add topics they want to see more often, remove ones they do not, and give feedback in plain language. Meta is also increasing its use of large language models in ranking, signaling a broader shift toward recommendation systems that can interpret user intent more like a conversation than a static settings page.

What Meta is changing on Facebook

Meta is extending a recommendation-control tool from Instagram to Facebook, giving users a more direct way to personalize their content mix. The company describes the new feature as a method for managing what appears in feeds and Reels, with updates intended to happen quickly after people submit feedback.

Facebook users will be able to access the controls in a few different ways. One option is through a dedicated Your Algorithm entry point. Another is by tapping “interested” or “not interested” from the three-dot menu on a post, a familiar interface pattern that now has a more explicit role in teaching the system what to prioritize.

Meta says the buttons may also show up directly below some posts, making the feedback loop easier to use without leaving the content itself. In some cases, the company says it may ask follow-up questions to get more detail about what a user liked or disliked.

Meta says people will be able to tell the system, in their own words, what they want to see more or less of, and that recommendations will update in real time based on that input.

How does Your Algorithm work?

Your Algorithm works by turning user preferences into immediate ranking signals that the recommendation engine can act on. In practical terms, that means someone who wants more posts about a topic, creator type or content style should see their feed adjust without waiting for long-term behavior to accumulate.

Meta says the feature can react within the same session, which is a meaningful change from the slower, more opaque feedback systems that have long defined social media feeds. Rather than relying only on passive signals such as watch time, clicks or likes, the platform is trying to make its algorithm easier to direct in real time.

Direct feedback, not just passive behavior

This update reflects a broader industry trend: platforms are moving from hidden personalization toward more visible user control. Instead of forcing people to infer how the feed works, Meta is offering a mechanism that lets them actively shape it.

That shift could matter for users who feel their feeds drift toward repetitive content or topics they no longer want. It may also reduce some of the frustration that comes with algorithmic feeds, where the path from user action to recommendation change can be hard to understand.

Why the timing matters

The Facebook rollout comes after Meta introduced similar controls on Instagram earlier this year and a comparable feature on Threads called Dear Algo. Bringing the idea to Facebook broadens the company’s test of whether users want to manage recommendations more explicitly across its platforms.

Because Facebook remains one of Meta’s largest products, the move could affect a much wider audience than the Instagram version did. If the tool gains traction, it may become a standard part of how people interact with Meta’s ranking systems.

Why Meta is leaning harder on large language models

Meta says it is increasingly using large language models, or LLMs, in its ranking systems because they can better understand meaning, context and user intent. The company’s pitch is that a model trained on language can interpret a request such as “show me fewer political posts” or “more home renovation videos” more naturally than older ranking systems.

According to Meta, LLMs can absorb the interests, context and intent behind natural-language input, which helps create what it calls an algorithm users can actually talk to. That language suggests the company sees recommendation ranking not just as prediction, but as an interaction.

In Meta’s framing, the advantage is responsiveness. When users state preferences directly, the algorithm can adapt in the same session rather than waiting for future behavior to confirm a pattern. That could make Facebook feel less like a black box and more like a system that can be steered on demand.

What LLMs add to ranking

Large language models are useful in recommendation systems because they can process more nuanced input than simple rule-based filters. They can infer whether a user is asking for a topic, a tone, a format or a creator style, even when the prompt is informal.

That does not mean the system will perfectly understand every instruction. But it does mean Meta is betting that natural-language feedback will make personalization easier for ordinary users and more useful for the platform.

How is Facebook’s version different from Instagram and Threads?

Facebook’s version is built on the same core idea as Instagram’s Your Algorithm, but it arrives in a different environment: a mature platform with a broader content mix, older user base and longer history of algorithmic friction. That makes the Facebook rollout potentially more consequential.

Instagram’s launch earlier in 2026 introduced the notion that users could more directly explain their preferences to Meta’s systems. Threads later added Dear Algo, a similar feature aimed at making recommendation tuning feel more conversational. Facebook now becomes the next major test case.

The key difference is scale and behavior. Facebook feeds are often used for news, family updates, local groups, creator content and Reels, so the company is applying the same tuning model to a more varied and politically sensitive stream of posts.

What this means for Facebook users

For everyday users, the change could make Facebook feel more responsive and less random. If someone sees too many posts about a topic they dislike, the new controls should make it easier to course-correct. If they want to explore a hobby, creator niche or theme, they can signal that preference more clearly.

That simplicity is important because most people do not want to study how recommendation systems work. They want a feed that reflects their interests with minimal effort. Meta is trying to lower the friction between intention and outcome.

• Users can add or remove topics from their recommendation profile.
• The system can react to “interested” and “not interested” feedback.
• Some posts may include visible buttons for quicker input.
• Meta may ask follow-up questions to refine the signal.
• Changes are supposed to update in real time.

At the same time, the feature also gives Meta more structured feedback data. The more directly people describe what they want, the better the company can train and adjust its recommendation systems. That is convenient for users, but it is also valuable for Meta’s product strategy and machine-learning pipeline.

Why recommendation controls matter now

Recommendation systems increasingly define the experience of major social platforms, but they are often criticized for being opaque, repetitive or overly optimized for engagement. Meta’s new controls arrive as pressure builds on platforms to give people more say over what algorithms surface.

By adding more transparent controls, Meta is responding to a user complaint that has lingered for years: people often feel trapped by feeds that overreact to a few taps or overexpose them to a narrow slice of content. A more interactive system may soften that problem.

It may also serve a strategic purpose. If users can self-correct the feed rather than leaving the app or muting topics manually, Meta could improve satisfaction while preserving engagement. That is a delicate balance, because the company still needs its recommendation engine to keep people scrolling.

The business logic behind the feature

From Meta’s perspective, more precise recommendation feedback can make rankings smarter and potentially more durable. Better signals can improve relevance, reduce user frustration and give the company another way to frame AI as a consumer benefit rather than only an infrastructure story.

There is also an industry-wide race to make AI feel useful in everyday products. Search, email, operating systems and social platforms are all experimenting with ways to let users steer models more directly. Meta’s update fits neatly into that broader push.

How the change fits into Meta’s AI strategy

Meta has been weaving AI into almost every part of its consumer products, from content ranking to assistants and creative tools. This Facebook update extends that effort into the core mechanics of discovery, which may be the most important AI use case the company has in front of billions of users.

Unlike flashy generative features that produce images or text, ranking AI operates quietly in the background. Yet it often has a bigger impact on daily experience because it determines what people actually see, watch and engage with.

That makes this rollout especially significant. It shows Meta is not only adding AI on top of Facebook, but using it to rework one of the platform’s foundational systems.

From black box to conversation

Meta’s language around the new system is revealing. The company is not just saying it uses AI; it is describing a feed users can “talk to.” That framing reflects a shift from passive personalization to interactive guidance.

Whether users embrace that approach will depend on whether the controls feel accurate, fast and easy to understand. If they work well, the feed may feel more personal. If they do not, the feature could become another settings menu that people ignore.

What we still do not know

Meta has not fully detailed how extensively the new controls will roll out, whether all users will see the same interface, or how much influence the feature will have relative to other ranking signals. It also has not explained whether certain kinds of content will respond faster than others.

Another open question is how consistently the system will interpret natural-language feedback. A request for “less drama” or “more local news” can mean different things to different people, and the strength of the model’s interpretation will likely affect user trust.

There is also the broader issue of algorithmic transparency. More control is helpful, but it does not fully reveal why content is recommended or how competing signals are weighted behind the scenes.

Timeline of Meta’s feed-control rollout

Meta’s new Facebook feature did not emerge in isolation. It is part of a sequence of product updates across the company’s social apps.

What comes next

The rollout will likely serve as a live test of whether people want to manage algorithms more actively or whether they prefer the system to do most of the work silently in the background. If engagement with the feature is strong, Meta may expand the model further across its apps.

For now, the update makes one thing clear: Meta is betting that the future of social feeds is not just smarter ranking, but ranking that can be directed by the user. In an era when AI is increasingly shaping what people see online, that kind of control may become a key selling point.

Facebook’s new feature is a small interface change with potentially large implications. It could help users regain some control over their feeds, while giving Meta a richer stream of preference data to improve its AI systems. In other words, the company is not only asking what people want to see — it is also teaching its algorithm how to listen.