Facebook's natural language algorithm controls will make topic consistency matter more than engagement bait

Facebook launched Your Algorithm on 13 October 2026, an AI-powered tool that lets users type their interests in plain language to directly shape what appears in Feed and Reels. Meta says the system adapts in real time, even within the same browsing session.

The feature provides a dedicated space where users can add and manage the topics they want to see, accessible through the three-dot menu on any post. After clicking Interested or Not interested, follow-up prompts ask users to explain what they want more or less of, with large language models processing the context and intent.

Meta first launched the feature on Instagram, before extending it to Threads (as Dear Algo) and now Facebook.

What it means for reach: topics you can name matter more than raw engagement

Organic reach on Facebook is decided by the systems this announcement changes, and until now distribution depended on what Meta inferred from behaviour such as what people stopped on, watched, liked or shared.

If users increasingly communicate specific interests directly to the platform, recommendation systems will have another source of information alongside observed engagement, although Meta has not disclosed the relative weight of these new inputs or quantified their impact on content reach.

The practical consequence is straightforward. Explicit user-stated topics now sit alongside engagement signals as ranking inputs, meaning content that does not match a user's named topics may surface less frequently regardless of its raw engagement rate.

For brands and creators, the distinction between content that attracts an initial interaction and content that remains relevant to a user's stated interests could become more significant. A post that generates comments but sits outside the topics a user has asked to see faces a distribution headwind that engagement alone may not overcome.

The change makes explicit audience interests part of the recommendation conversation, but does not provide a new marketer-controlled targeting field, and Meta has not supplied evidence that the feature will increase a brand's reach.

The algorithmic shift Meta isn't saying out loud

By the second quarter of 2026, Meta was describing a Meta Generative Recommender that uses large language models to evaluate ad content and user preferences together, and reported that early pilots using LLMs to understand user preferences produced a 1% lift in app event conversions on Instagram.

This update brings that LLM layer into organic ranking. Meta says large language models read the meaning and intent in what users type and combine it with the thousands of signals Facebook already tracks.

What Meta frames as giving users control is also a data collection exercise. Every topic a user names, every follow-up answer they provide, trains the model to understand preference in language rather than just behaviour. That makes the algorithm better at predicting what someone wants before they interact with it, which compounds the advantage for content that fits cleanly into the categories users have already articulated.

Consistency beats virality when users can suppress your category

Instagram launched Your Algorithm in December 2025 for Reels, then expanded it to Explore and eventually the main feed, showing users an AI-generated summary of the topics influencing their recommendations. Instagram has been introducing it to more areas of the app, with head Adam Mosseri saying the goal is to evolve Your Algorithm from a setting to something central to the Instagram experience.

Brands running creator campaigns on Instagram have already felt the effect. Briefs that instruct creators to stay tightly within a product vertical now risk being invisible to users who have suppressed that category; the content looks correct but delivers to a shrinking audience.

The fix is not to abandon relevance. It is to brief creators on narrative framing that spans adjacent topics; a skincare brand whose core audience might suppress beauty still lands if the creator frames the content around morning habits, self-care routines or productivity rituals, topic clusters less likely to be suppressed.

On Facebook, the same logic applies. If your brand sits squarely in a category users can easily name and suppress (fashion, beauty, finance, tech gear), every piece of content you publish now competes not just for attention but for inclusion in a preference list the user may never have added you to.

The answer is not to obscure what you do. It is to understand which adjacent topics your audience has asked to see and to frame your content accordingly. A DTC furniture brand talking only about products will lose reach to one that talks about interior design, small-space living and home productivity, because those are the terms people actually use when they describe what they want to see more of.

What changes on Monday

Three practical shifts for anyone trying to reach people on Facebook:

Test whether your content maps to topics people ask for. Run a content audit. For each piece, write down the topic a user would need to have added to Your Algorithm for that post to reach them. If you cannot name it clearly, the LLM probably cannot either.

Stop optimising for engagement bait if it lives outside your core topic. A post that generates comments through a generic "tag someone who needs this" hook may score well on engagement but poorly on topic relevance. If the user never asked to see that type of content, the new system will route it away from them regardless of how many people interacted with it.

Assume reach will fragment by declared interest, not just inferred behaviour. Until now, you could reach someone by mimicking the signals of content they engaged with. Now you also need to match the language they used to describe what they want. If your content doesn't fit a topic someone has named, you are starting from a weaker position than you were a month ago.

Meta has not disclosed how much weight the new inputs carry relative to behavioural signals, nor has it quantified the impact on content reach, and has not announced any specific advertising product tied to Your Algorithm.

Meta described the feature as rolling out but provided no details on which countries would receive it first or when broader availability would follow.

The strategic question: do you own a topic users will name?

The update accelerates a trend already visible across Meta's broader algorithm changes and in how other platforms are shifting toward user-controlled feeds.

Reach is moving away from passive inference (the platform guesses what you like based on what you watch) and toward active declaration (you tell the platform what you want and it delivers that). The beneficiaries will be accounts and content that map cleanly onto topics users actually ask for. The losers will be anyone relying on virality, broad appeal or engagement tactics that sit outside a nameable interest category.

If your brand strategy is built around being famous rather than being useful in a specific, easily articulated domain, this shift will hurt. If you have been coasting on engagement hooks that do not reinforce what you are known for, the algorithm will now penalise you for it.

The question every brand and creator should answer is simple: when someone opens Your Algorithm and types what they want to see, is your content what they are asking for? If the answer is not obvious, your reach problem just got worse.

For context on how Meta's algorithm has shifted creator and brand reach over the past year, see our analysis of Instagram's single-take content boost and why follower engagement is now a reach liability on TikTok, patterns that are now extending to Facebook.