Your algorithm on Facebook: what Meta revealed and what it means for creators
Meta published how Facebook's Your algorithm control works under the hood: a new topic gets a temporary boost, every topic has a daily cap, and asking for less of a creator filters their content out directly, with no gradual downranking. Here is what that changes for anyone posting to Feed and Reels.
Your algorithm on Facebook: what Meta revealed and what it means for creators
On November 10, Meta published an engineering blog post explaining how Facebook's Your algorithm works under the hood. It is the panel where users say in plain words what they want more or less of in Feed and Reels. For anyone who posts, three details matter more than the panel itself: a newly added topic gets a boost for a limited time, every topic has a daily exposure cap, and when someone asks for less of a creator, they stop getting that creator's content through a direct filter, not a downranking.
The post does not announce a new feature. The control launched on October 13, as TechCrunch reported. What is new is that Meta opened up the rules for how a user request becomes a recommendation, which helps explain why the reach of a single profile swings so much from one video to the next.
How does the panel work for users?
Users open the panel from the three-dot menu on any post in Feed or Reels. There they find a sentence summarizing what they have been watching lately, a list of topics of interest that Meta has already filled in, and a list of topics they have turned down.
They can type a new topic in their own words, delete one, or move it from one list to the other. Tapping "Interested" or "Not interested" on a post also feeds these lists, and Facebook may ask follow-up questions about what the person did or did not like. Meta says recommendations change within the same session.
On Instagram, the same control arrived earlier: on Reels in December 2025, and on the feed and Explore in June 2026, according to TechCrunch. We have already covered what it demands from your video on Instagram. Facebook follows the same logic, but now with the mechanism described by the company itself.
About Brazil: Meta talks about a gradual rollout and has not published a list of countries. There is still no confirmation that every Brazilian account already has the panel.
What happens behind a request?
A language model reads the person's interest signals, which are updated often, and turns them into readable topic names. When the user types a topic, a smaller model checks that it is a coherent subject, merges duplicates and gives it a clean name. After that, the topic gets its own sources for finding content, and videos that match it receive a promotion in the final ranking stage.
Four rules in Meta's post deserve a close read:
- A new topic weighs more for a limited time. In Meta's words, "new interests get extra weight for a limited period." After that, what decides is how the person reacts to what they received.
- No single topic takes over the feed. There is a layer that spaces out recommendations on the same topic and limits how much each topic contributes per day. The company's line: "a new interest should show up, not take over."
- A quality request is not a filter. Asking for less AI-made or sexually suggestive content triggers dedicated downranking signals, not the topic logic. The content drops in the ranking, but it does not disappear.
- A request about a creator is a filter. "When a request names a creator, that creator's content is filtered directly." No gradation.
For rejected topics, filtering happens in several stages: in the ranking, in a model that reviews the final list and removes anything matching a recent "not interested," and even on the phone, deleting videos that were already queued to load.
Who wins and who loses?
The winners are creators with a clear subject and an audience that comes back on its own. The losers are those who live off generic videos that hitch a ride on any topic, and videos that annoy people to hold their attention.
A video with no topic misses the promotion. The promotion in the final stage goes to videos that match a requested topic. If the machine cannot tell what your video is about, it does not get that advantage. In clips, the classic problem is the excerpt that starts on a reaction ("dude, look at this") and never says the name of the game, the guest or the subject.
The daily cap changes the competition. When someone adds "soccer" and Facebook limits how much soccer shows up per day, every soccer profile competes for the same small quota in that person's feed. Being on the topic is not enough: within the quota, whoever holds attention wins. The ranking signals are still the usual ones, such as likes, comments and watch time, according to Meta. We explained the weight of sending by message in why a send is worth more than a like.
The new-topic boost explains a spike followed by a drop. If your video reached a wave of people who had just added your subject, it gets a surge that does not last. When the extra weight ends, only real engagement is left. Many videos that "died out of nowhere" on day three may be this.
AI content pays a quiet price. Someone who ticked "less AI" will not stop seeing your generated video, but it will show up less for that person. Meta does not say how it identifies what is AI-made in this case. The platforms' trend of rewarding what looks human-made is covered in the post on generic AI content on YouTube, and Facebook is heading the same way.
Blocking by creator is the costliest risk. Clickbait, misleading titles and tiring repetition have always cost retention. Now they can cost you the whole person, for as long as the request stays on their list, with one sentence typed into the panel.
Why is Meta opening this up now?
Part of it is product: the company has been crediting recommendations for growth in time spent. According to Social Media Today, in its second-quarter 2026 earnings Meta said video time on Facebook rose 9% worldwide, driven by improvements to the feed and Reels recommendations.
Another part is regulatory pressure. The same outlet notes that Australia began studying rules that would let people turn off algorithmic recommendations and make the algorithm-free feed the default. In the outlet's reading, Meta hopes the visible control will cool the push to bring back the chronological feed. And Social Media Today itself points out that few people touch these settings.
What I would do starting today
My take: only a minority will use the panel itself, but the rules described apply to everyone, because the "Interested" and "Not interested" buttons in the three-dot menu feed the same system. So I would treat this as a distribution rule, not a settings curiosity.
- Say the subject in the first seconds, spoken and written on screen. Name of the game, the guest, the topic. That is what makes the video fit a requested topic.
- Write the post description naming the subject in the first line, without relying on emoji and loose hashtags.
- Cut the bait that annoys people. A title that promises what the video does not deliver can now get your profile blocked, not just skipped.
- Do not draw conclusions from a single video. A spike and a drop within three days may be the temporary topic boost running out. Look at the average of ten videos.
- If you use AI, use it for what improves the video, not to fabricate the whole video with a generated look. Anyone who asked for less AI will see you less.
For anyone clipping livestreams and podcasts, the first item has the best return: choose the entry point where the subject is stated, not the reaction. On Cut.Pro, clips come out with a transcript and burned-in captions, you can adjust where the excerpt starts before publishing, and you can schedule to Facebook along with the other networks (TikTok, Instagram, Facebook, Threads, X, LinkedIn, Bluesky, Pinterest, Telegram, Mastodon, Discord, Dailymotion, Tumblr, Slack, Car.Media, and YouTube). If Facebook has become an important channel for you, it is also worth reading how to build a clip channel on Facebook Reels and what Creator Fast Track pays.
Meta still has not published signal weights or a formula. What it opened up is enough for a practical conclusion: a video that says right away what it is about and does not annoy the people watching is what this system favors.
