For many years, people have influenced social-media recommendations through indirect signals: watching longer, liking, commenting, skipping, or choosing “not interested.” Those signals can be useful, but they do not always express what a person actually wants. Someone may watch a video out of curiosity without wanting more of the same. Conversely, a new need may not yet have generated enough behavioural history for a system to notice it.
Fact: on 13 October 2026, Meta announced new AI-powered tools for Facebook, including “Your algorithm,” which lets people add and manage topics they want to see in Feed and Reels. Meta also says that after people select “Interested” or “Not interested,” they may be asked follow-up questions and can describe in their own words what they want to see more or less of; the company says recommendations will adjust in real time within the session. These are Meta’s statements about its features and systems, not independent evidence that every user will receive the same result.
From behavioural signals to stated intent
The notable part of the announcement is not only a new button. It signals a shift toward recommendation systems combining behavioural history with intent expressed directly by the user. Meta describes using large language models to bring interests, context, and intent from free-form text into ranking. If implemented as described, users may be able to correct a system’s inference more quickly when their interests change.
This does not mean an algorithm will understand every need. Stated intent can still be brief, ambiguous, or situation-dependent. For marketers, however, it is a reminder that an impression or a view does not always equal sustained interest. Clearer signals of value, relevance, and a reason to keep watching are more useful for a team to monitor.
What it means for brand content
When users have more ways to say what they do or do not want to see, content that relies only on a curiosity gap may encounter a clearer limit. A headline can generate an initial click, but if the content does not fulfil its promise, viewers have more ways to reduce that type of content in their experience. This is not a reason for brands to avoid engaging topics. It is a reason to keep the headline promise, the actual content, and the call to action consistent.
For example, a health-and-wellness brand might create a video about sustaining an exercise habit. Instead of using a broad claim such as “the secret to changing your life,” it can state that the advice is for office workers who want to move for ten minutes during the day. The right viewer can recognise the value; an unsuitable viewer is not drawn into a different expectation. That clarity helps create higher-quality feedback, however the platform measures it.
Design content so viewers can choose for themselves
A better recommendation system does not replace the need to understand customers. A brand still needs to know whom it serves, in which circumstance, and with what specific benefit. Content can support this by stating who it is for, the use situation, and the limits of the advice. These details do not make content less engaging. They help viewers decide more quickly whether it deserves their time.
A B2B business, for instance, could divide an AI series into episodes for operations managers, sales leaders, and data administrators. Each episode begins with a concrete job, such as reducing reporting time or checking input quality. When viewers select content for their work, engagement is more likely to reflect relevance than random curiosity.
Measure response quality, not only reach
For a marketing team, Meta’s announcement suggests a practical question: which signals show that content is genuinely useful? Reach still matters, but it should sit alongside deeper signs such as meaningful watch time, saves, specific questions, return visits, movement to related content, and negative feedback. No single metric replaces all others; what matters is distinguishing temporary attention from signs that viewers want to build on the relationship.
Set one simple standard for each content type. For an explanatory article, it might be the number of readers continuing to a related resource. For a tutorial video, it might be completion rate and application questions. For a product introduction, it might be the number of viewers who move to relevant information rather than merely skim. Teams should review these measures over time and by content group, not infer long-term effectiveness from one spike.
Management view
Meta giving users a more direct way to speak to a recommendation system does not make marketing a game controlled entirely by the platform. It reinforces an older but increasingly important principle: content has value when it helps the right person understand or do something specific. As user feedback becomes clearer, the gap between content that attracts attention and content that people want to keep seeing may become easier to observe.
Businesses do not need to wait for every new feature before acting. They can begin by reviewing their ten most recent pieces: what does each promise, who is it for, what can the viewer gain in the first few minutes, and which signal would show that it was genuinely useful? This helps a brand build content on relevance and trust rather than optimise only for a single appearance in a recommendation feed.
