How the X (Twitter) Algorithm Works: What Creators Can Control

X’s algorithm decides which posts appear in your For You timeline, which replies show first, and how far a post travels beyond the author’s followers. X has explained the broad strokes: it mixes posts from accounts you follow with recommended posts from outside your network, using signals such as popularity and how your network interacts. It doesn’t publish the full ranking formula, and it changes over time.

So the useful question isn’t “what’s the secret formula?” It’s “which signals can I actually influence?” This guide covers what X has said publicly, what creators commonly observe, what’s myth, and a practical playbook built on the parts you control.

Two timelines: For You and Following

When you open X, you see two main tabs:

	For You	Following
Whose posts	Accounts you follow plus recommended accounts you don’t	Accounts you follow
How it’s ordered	Ranked by X’s recommendation system	Generally the most recent posts first
What it means for creators	Your chance to reach non-followers	Your chance to reach existing followers reliably

For creators, For You is where growth happens, because it’s the main place a stranger can meet your post. Following is where loyalty shows: the people who chose to see you.

In-network vs out-of-network

The most useful mental model is to split your audience in two:

• In-network: people who follow you. Your posts can appear in their Following tab and, often, in their For You tab.
• Out-of-network: everyone else. They see your post only if X recommends it, someone they follow reposts it, or they find it in search or on your profile.

According to X’s help center, X adds posts from outside your network using signals that include how popular a post is and how people in your network interact with it. X describes the aim as showing content that is relevant, credible and safe.

In practice, this means a post usually has to do well with the people closest to it before it travels further. If your followers — and people similar to them — respond, X has a reason to show it to more of their neighbors.

What X has said publicly

A few details come straight from X’s help center, and they’re more actionable than most algorithm theories:

1. Recommendations use popularity and network interactions. Engagement from people connected to the viewer matters.
2. The goal is relevant, credible and safe content. Posts that may break the rules can be shown less.
3. Premium replies get a slight preference. X says replies from verified (Premium) accounts get a “slight preference” in conversation ranking.
4. Posts that begin with @username are limited. They appear in the Home timelines of people who follow both accounts — not all your followers.
5. Replies have their own distribution. A reply shows in the Home timeline of people who follow both the sender and the recipient, or if X thinks it’s relevant.

In 2023, X also published parts of its recommendation code publicly. That release gave outsiders a rough view of how candidate posts are gathered and ranked, but it wasn’t a live rulebook, and systems like this are adjusted continually. Be skeptical of anyone quoting precise “weights” as current fact.

Signals creators can influence

You can’t see X’s formula, but you can control the inputs that plausibly feed it:

Signal	What you control
Early engagement	Post when your followers are online; write a first line worth stopping for
Replies and conversation	Ask real questions; reply to early commenters quickly
Relevance to your audience	Stay on the topics people followed you for
Media engagement	Use native images and video that earn clicks and watch time
Credibility and safety	Avoid spammy tactics, misleading claims and rule-breaking content
Network effects	Build relationships with accounts whose followers overlap with yours

A good way to see these signals is your own analytics. Impressions show how widely a post was shown; engagement rate (engagements ÷ impressions) shows how well it landed. If you’re new to those numbers, read what impressions are on X and what a good engagement rate on X is.

Things you can’t control (and shouldn’t obsess over)

• The exact ranking weights. X doesn’t publish them, and they change.
• What each viewer has engaged with before. Two people see different For You timelines.
• The news cycle. A major event can crowd out almost everything for a day or two.
• Tests X is running. Platforms regularly experiment, which can make reach feel inconsistent.

Treat unexplained week-to-week variation as noise. Look for patterns across a month of posts instead. If reach falls and stays down, work through the checklist in why your X impressions are dropping.

Algorithm myths vs reality

Myth	What’s more likely true
“Hashtags boost reach.”	X doesn’t say they do. One or two relevant tags can aid search; many look spammy.
“Links are banned from reach.”	Links aren’t banned, but many creators report lower reach on link posts. Test a link-in-reply version.
“Editing a post kills its reach.”	X keeps engagement on edited posts and shows an “edited” label. There’s no public evidence editing is penalized.
“Premium guarantees reach.”	X describes a slight preference for Premium replies in conversations — not guaranteed distribution.
“Posting 20 times a day is the key.”	Volume helps only if quality holds. Weak posts drag down your averages, and X has posting limits.
“Deleting and reposting resets the algorithm.”	It wipes your data and can look like duplicate content.
“Any drop means a shadowban.”	Most drops come from timing, format, topic or engagement changes. See our shadowban guide.

A practical playbook

Here’s how to work with the system instead of guessing at it:

1. Know your baseline. Track median impressions and engagement rate for your last 20–30 posts.
2. Win in-network first. Post when your followers are active and write for them. Out-of-network reach tends to follow strong in-network response.
3. Make the first line do the work. Lead with the claim, number or question.
4. Start conversations, not announcements. Posts that invite replies give X more interaction to work with — and give you more to reply to.
5. Reply to your repliers. Early back-and-forth keeps a post active while it’s being evaluated.
6. Use native formats. Images, native video, polls and threads keep people on the post. Our guide to writing a thread on X shows how to structure one.
7. Don’t start broadcast posts with @username. Put a word first so all your followers can see it.
8. Reply thoughtfully in your niche. Good replies under relevant posts are how many accounts get discovered.
9. Avoid anything that looks like spam. Mass-tagging, repetitive text, aggressive automation and bought engagement all work against the “credible and safe” goal X describes.
10. Review monthly. Double down on what beats your median; drop what doesn’t.

Key takeaways

• For You mixes posts from people you follow with recommendations from outside your network; Following shows accounts you follow.
• X says it recommends posts using signals including popularity and how your network interacts, aiming for relevant, credible and safe content.
• X published parts of its recommendation code in 2023, but that isn’t a current, exact formula.
• Focus on what you control: timing, first lines, conversation, native media, topic focus and clean behavior.
• Judge the algorithm by your own trends over weeks, not by a single post.