How do social media algorithms decide what to show you?

A ranking model scores a pool of eligible posts based on predicted engagement — using signals like past likes, comments, shares, and watch time — then sorts and displays the highest-scoring results first, personalized to each account.

Why do two people see different things in their feed even if they follow the same accounts?

The ranking model scores the same pool of candidate posts differently for each person based on their own individual engagement history, so identical follow lists can still produce very different personalized rankings.

Do social media feeds show posts in chronological order?

Usually not by default — most platforms rank by predicted engagement rather than posting time. Some platforms offer an optional chronological feed setting that bypasses the ranking step entirely.

Why does misleading content sometimes spread so widely on social media?

Ranking models typically optimize for predicted engagement, and strong reactions — including outrage or disbelief at inaccurate claims — register as high engagement regardless of whether the content is true, which can help it spread quickly.

Can you turn off the algorithm on social media?

Some platforms offer a chronological or 'following only' feed setting that bypasses personalized ranking, though the underlying platform may still use algorithms elsewhere (like search or recommendations outside the main feed).

How Social Media Feeds Decide What You See — KnowThisWell

Key Takeaways

1A feed's order comes from a ranking model scoring a narrowed-down pool of candidate posts against predicted engagement — not strict chronological order, and not a human editor's pick.
2The model doesn't judge truth, importance, or quality — it predicts what you're statistically likely to react to, based on your own past behavior and similar users' behavior.
3Your reactions to what you're shown feed back into the same model, which is why two people with different habits on the same platform see very different feeds even for the same topic.

The concept

Picture a librarian who, instead of shelving books alphabetically, hands you the books she thinks you personally are most likely to actually read, based on every book you've picked up before. A social media feed works similarly: instead of showing every post from everyone you follow in the order it was posted, the platform predicts which posts you're most likely to react to and puts those first. What you see is a personalized prediction, not a neutral list.

Once a feed is understood as a prediction, not a neutral timeline, patterns that used to feel arbitrary — why certain posts blow up, why two people's feeds diverge so sharply — start making mechanical sense.

Quick check

Two friends follow all the same accounts on a social platform, but their feeds look completely different. What best explains this?

Worked examples

Example 1: Liking posts about one topic changes what shows up next (baseline case)

Someone starts liking and lingering on cooking videos on a social platform. Within a short time, their feed shifts to show noticeably more cooking-related content, even from accounts they don't follow. This is the feedback loop completing a full cycle: the model logged the engagement signals (likes, watch time) as evidence of interest, updated its prediction for that person, and began surfacing more candidates it scored as similarly likely to engage. Nothing was manually curated — the shift is a direct, mechanical response to the new engagement data.

Quick check

Someone notices their feed shifted toward a new topic shortly after they engaged with a few posts about it. What caused the shift?

Example 2: A widely shared post that turns out to be misleading (edge case / variation)

A post containing inaccurate information spreads rapidly because it provokes strong reactions — outrage, disbelief, heated comments — all of which register as high engagement to a ranking model. The model has no built-in concept of factual accuracy; it optimizes for predicted engagement, and strong emotional reactions are a powerful engagement signal regardless of whether the underlying claim is true. This is a well-documented structural property of engagement-based ranking, not a bug specific to one platform, which is why many platforms have added separate systems — fact-checking labels, reduced distribution for flagged content — layered on top of the base ranking model specifically to counteract it.

Quick check

Why can a misleading post sometimes get ranked and shown very widely by a feed algorithm?

Example 3: Understanding why a switch to chronological order changes the experience (real-world / applied case)

Some platforms offer a setting to switch from the default ranked feed to a strictly chronological one. Choosing chronological order effectively removes the ranking model's scoring and sorting step entirely, showing posts purely in the order they were posted from accounts a person follows. The tradeoff becomes visible immediately: chronological feeds show far less content overall (since there's no candidate-narrowing step optimizing for relevance) and can surface far more low-engagement or off-topic posts, but they also stop reinforcing the feedback loop that narrows a person's exposure toward whatever they've engaged with most.

Quick check

Someone switches their feed setting from the default ranked view to a chronological view. What actually changed under the hood?

How it works (visual)

From candidate pool to personalized feed

Common mistakes

Common Mistakes

Assuming a social media feed shows posts in the order they were made.

→ Most default feeds are ranked by predicted engagement, not posting time — a strictly chronological feed is usually a separate, optional setting where available.

Believing the algorithm knows or cares whether content is true.

→ Base ranking models optimize for predicted engagement, not accuracy — factual verification, where it exists, is typically a separate system layered on top.

Thinking your feed is identical to what other people following the same accounts see.

→ Ranking is personalized per account based on individual engagement history, so identical follow lists can still produce very different feeds.

Assuming the feed is a fixed, unchanging list rather than a live prediction.

→ The ranking updates continuously as new engagement data comes in — the same post can rank differently for the same person at different times.

Common misconception

“Social media feeds are just a neutral, chronological list of what people you follow have posted.”

Most default feeds run every eligible post through a ranking model that predicts engagement and sorts accordingly — the order is a personalized prediction, not a neutral timestamp sort. This isn't inherently deceptive; it's a deliberate design choice to surface what a model estimates you'll find worth your time. But it does mean the feed reflects a statistical guess about your preferences, filtered through what the platform can measure, not an unfiltered record of everything posted.

What to do next

If a platform offers a chronological feed option and you want less algorithmic filtering, try switching to it and compare what you see.
Remember that liking or lingering on a type of content will likely shift future recommendations toward more of it — that's the feedback loop, not a coincidence.
Be more skeptical of content that provoked a strong emotional reaction and spread quickly — that pattern is exactly what engagement-based ranking tends to amplify, regardless of accuracy.
Read How Push Notifications Work next to see the related mechanism that pulls you back to check a feed in the first place.

FAQ

Related terms

Ranking model

Candidate selection

Engagement signal

Feedback loop

Key Takeaways

1A feed's order comes from a ranking model scoring a narrowed-down pool of candidate posts against predicted engagement — not strict chronological order, and not a human editor's pick.
2The model doesn't judge truth, importance, or quality — it predicts what you're statistically likely to react to, based on your own past behavior and similar users' behavior.
3Your reactions to what you're shown feed back into the same model, which is why two people with different habits on the same platform see very different feeds even for the same topic.