How social media feeds work

For definitions of the two terms on their own, see algorithmic feed and chronological feed.

1. 1

   Gather candidates

   The system collects posts you could be shown: recent posts from people you follow, and on many platforms posts from accounts similar to ones you engage with.

2. 2

   Score each candidate

   Each post gets a score from signals about the post, its author and you. Big platforms use machine-learning models that predict actions such as a like, a comment or watching a video to the end. Smaller apps often use simpler weighted formulas.

3. 3

   Apply rules on top

   The sorted list is adjusted: remove what breaks the rules, avoid showing five posts in a row from one account, mix in different formats, keep things reasonably fresh.

4. 4

   Learn from what you do

   What you react to, skip or hide feeds back into the next ranking. This loop is why a feed can drift toward one topic surprisingly fast.

Every platform weighs things differently and few publish exact formulas, but the ingredients are well understood.

• Engagement on the post — reactions, comments, shares and saves. Comments and shares usually count as stronger signals than a quick like. See what engagement means.
• Recency — newer posts get a boost that fades over hours or days.
• Your relationship with the author — whether you follow them, and how often you interact.
• Your history — the topics, formats and creators you've engaged with before.
• Watch behavior for video — whether people finish it, rewatch it or swipe away in the first seconds.
• Negative feedback — hiding, muting, reporting, or consistently scrolling past.
• Integrity signals — spam patterns and rule violations, which push content down or out.

“Why am I seeing a post from two days ago?” It had enough engagement to outscore newer posts. In a ranked feed, age is one signal among many.

“Why didn't my friend's post show up?” Ranked feeds show a slice, not everything. If you rarely interact with that friend, or the post got little early activity, it may have ranked below the point where you stopped scrolling. A chronological view would have shown it.

“Why is my feed suddenly all one thing?” The feedback loop. Linger on three cooking videos and the system reasonably offers a fourth. Hate-watching counts too — the feed sees attention, not approval.

“Why does one account dominate?” A prolific account with high engagement can crowd everyone else out unless the feed limits how often one author appears. Many feeds do; not all.

Ranking has a built-in blind spot: it relies on history, and new things have none. A brand-new account has no engagement record to learn from, and a post published a minute ago has no reactions yet. This is called the cold-start problem, and every ranked feed needs some answer to it.

• For new accounts, apps usually ask for interests at sign-up, suggest people to follow and fill the gaps with broadly popular content until they know more. Asyncbit's onboarding, for example, asks you to pick interests from 16 topics, and its people suggestions draw on shared interests and mutual follows.
• For new posts, feeds often show a fresh post to a small group first and let early reactions decide whether it travels further. Recency weighting does part of this job automatically.
• For new creators, some feeds add an explicit boost for content with few views, so it gets a fair first audience.

This is also where the “are feeds designed to be addictive?” question lives. A feed tuned only to maximize engagement will tend to reward whatever grabs attention fastest, including outrage, and a feed with no end has no natural stopping point. A chronological tab, limits on how often one author appears and meaningful mute tools are the counterweights to look for in any app you use.

Asyncbit's social feed keeps the two models separate and lets you pick, which makes it a useful illustration.

A few things it deliberately doesn't do: there is no trending-posts feed, only trending hashtags; you browse a hashtag page (newest first) rather than following tags; and the Following tab never inserts suggested posts. The new-reel boost addresses a known weakness of ranking short-form video purely on engagement: a video with no views has no engagement yet, so it would never get the chance to earn any.

1. Use the chronological view when completeness matters. If you need to see everything from specific people, switch tabs rather than hoping the ranking surfaces it.
2. Engage on purpose. Comment on and save what you want more of. Scroll past what you don't — and don't argue in the comments of posts you dislike, because that is engagement.
3. Mute and unfollow freely. Muting is the strongest negative signal most users have, and it is invisible to the other person.
4. Browse deliberately. Search and hashtag pages give you what you asked for, not what the ranking guessed.
5. Reset drift. If your feed has narrowed, spend a few sessions engaging with a wider mix and hiding what you're tired of.

In an engagement-and-recency feed, the first hour after posting matters because recency fades while engagement builds. Post when the people you want to reach are around. In a chronological feed, timing is almost everything — a post at 3 a.m. is buried by morning.

Don't bother flooding. Feeds with an author-diversity guard spread your posts out, so ten posts in an hour mostly compete with each other. Engagement bait (“comment YES if…”) tends to be spotted by readers, and increasingly by ranking rules too. The reliable route to reach is still posts people actually want to reply to — and a real audience, which you build by finding people who share your interests and taking part in their conversations.