Facebook Lets People Tell Its AI What Belongs in Their Feed

Facebook users can now describe the topics they want in Feed and Reels, following Meta’s October 13 announcement. For a marketer publishing there, a viewer’s written preference becomes another input that can shape whether the next post reaches them.

The feature, Your algorithm, sits in the three-dot menu. Meta says its AI ranking can adapt within the same session. That’s a product capability claim, with no accompanying experiment showing higher brand reach or more sales.

In its announcement, the company describes the mechanism as:

“by bringing freeform text together with everything we already understand about what you enjoy”

A person can now correct the interests the system infers from their behavior. The practical question for a publishing team is how clearly a post serves an interest that person wants to pursue.

The viewer adds an instruction

Until now, Facebook’s ranking systems inferred preferences from engagement, including likes, comments, and viewing. Meta says the new text input works alongside those existing signals. It hasn’t announced their replacement or published the weight assigned to a person’s written request.

People can manage topics directly. After tapping Interested or Not interested on a post, they may also receive follow-up questions and describe what they want more or less of. Large language models, the AI systems that interpret ordinary written requests, help the ranking system understand the intent behind those answers.

Those controls belong to the viewer. The announcement doesn’t establish a corresponding field that an advertiser can target, or explain whether the requests affect ad selection. An organic publishing team has a concrete interface to examine; a paid team has no basis here for changing an audience-targeting plan.

The distinction matters for a small B2B company, too. A post written for an existing customer often assumes the reader already knows the product. A recommendation can reach someone without that background. Reviewing whether the subject and intended audience are clear is a useful editorial check, even before a team can measure the effect of these new controls.

Narrower interests can change the publishing brief

Lindsey Gamble is now vice president of creator strategy and innovation at IZEA. When Instagram introduced its own Your Algorithm controls, they considered the audience implications in a December 11, 2025 newsletter:

“as personalization becomes more user-driven, platforms will inevitably fragment even further.”

That was an assessment of Instagram and the broader direction of social recommendations, months before this Facebook launch. It’s a forecast about audience fragmentation, rather than a measured result from either platform.

For marketers, it suggests a reason to compare topic-specific content with broadly framed brand posts. A useful comparison would hold the format reasonably consistent and record what each post was about alongside its results. This is a proposed test, not evidence that a narrower topic will win.

The writing brief also needs restraint. Meta hasn’t supplied a creator instruction saying that particular words or captions guarantee matching to a chosen interest. Adding a list of topics to every post would turn an untested interpretation of the feature into a production rule. Begin with the viewer’s understanding of the post, then check what the account’s results show.

Our reporting on AI product imagery and brand trust raised a separate creative question: what people accept when they encounter brand content. These recommendation controls add a distribution question. Reaching an interested person and persuading them remain separate jobs.

Access to a control doesn’t establish that people use it

Rachel Karten, the social media consultant who writes Link in Bio, offered a skeptical view of Instagram’s controls in a June 11, 2026 newsletter:

“I’m not sure if users really appreciate a “solution” to a problem that they perceive Instagram as the creator of.”

That older observation questions the appeal of the interface, not its technical operation. It’s useful counterweight to assuming every viewer will actively maintain a list of interests. Facebook hasn’t released adoption data or a geographic rollout schedule in this announcement.

It also keeps this story distinct from Meta’s recent Muse assistant changes. Here, AI interprets a person’s request to adjust recommendations. It isn’t carrying out a shopping task or managing a marketer’s campaign.

For the next Facebook publishing review, add a topic field to the content record and check whether each post makes that topic understandable without prior brand knowledge. Preserve the existing performance comparison. Keep any claimed effect of Your algorithm out of the report until the evidence supports it.