How Social Media Algorithms Decide What You See

There is no person choosing what appears on your screen — but there is a prediction behind every post. Once you understand what is being predicted, social media stops feeling random.

Why feeds aren't chronological

The average user could see thousands of new posts every day. Nobody can scroll through all of them, so platforms rank them. For each candidate post, models estimate: will you stop and watch? Share it? Save it? Comment? Skip it? Then they sort by a weighted mix of those predictions.

Be the algorithm

Change the weights or try a preset and watch which post wins. Notice what happens to the “link in bio” promo.

The signals that matter

• Watch time & completion — did people stop scrolling, finish, rewatch?
• Shares / sends — sending a post to a friend is a powerful signal of value.
• Saves — “I want this later”: tutorials, lists, recipes.
• Comments & replies — conversations, especially between real people.
• Relationship — accounts you interact with get priority in your feed.
• Negative signals — “not interested”, quick skips, hiding, reports.

What creators should do

1. Win the first 3 seconds — see the retention curve.
2. Make it sendable: relatable, useful, or surprising enough to share.
3. Make it saveable: checklists, steps, before/after.
4. Spark real comments: ask a specific question, not “thoughts?”.
5. Be consistent in topic so the system knows who to show you to.

“Don't make content the algorithm likes. Make content people send to a friend — the algorithm follows people.

Myths

“Shadowban for using a link”

Links that send people off the app often get less reach because people leave — not necessarily a punishment, just weaker signals.

“More hashtags = more reach”

Today hashtags are mainly a topic hint. A few relevant ones help; stuffing doesn't.