Algorithm

How the X Algorithm Actually Works (Updated Jan 2025)

A technical breakdown of X's recommendation algorithm and what it means for your content strategy.

Jan 10, 2025 12 min read TweetGoat Team

X open-sourced its recommendation algorithm in 2023, and it has evolved significantly through 2024. Understanding how it works gives you a real competitive edge. Here's the current state of the algorithm and what it means for your strategy.

The Three Phases of Content Ranking

Every tweet goes through three phases: Candidate Generation (the algorithm selects ~1,500 candidates from your network and beyond), Ranking (each candidate is scored using a neural network with 48K+ features), and Filtering (diversity and safety filters are applied). Your goal is to score well in the Ranking phase.

Flowchart: Candidate Generation → Ranking (48K features) → Filtering → Your Timeline
How the X recommendation pipeline processes every tweet before it reaches your timeline.

The Engagement Score Weights

Not all engagements are equal. Based on the open-source code and confirmed by our testing:

Key Stat

27×

A reply is weighted 27x more than a like in the X algorithm

Key Info

Engagement weights: Replies = 27× | Profile clicks = 12× | Dwell time (2+ min) = 10× | Bookmarks = 4× | Retweets = 1× | Likes = 1× (baseline). Optimise for replies and bookmarks above all else.

What This Means For Your Strategy

Optimise for replies above all else. Write tweets that invite conversation. Ask questions. Make contrarian statements. The second priority should be bookmarks — create content so useful people save it for later. Long-form threads that hold attention drive dwell time, which is heavily weighted.

PS
Priya Sharma@priya_grows

People spend hours optimising for likes. The algorithm barely cares about likes. Here's what it actually weighs: → Replies: 27x → Profile clicks: 12x → Dwell time: 10x → Bookmarks: 4x → Likes: 1x Write for replies. Everything else follows.

2:30 PM · Mar 26, 2026

2035783124891

The Network Effect

The algorithm heavily favors content from accounts you regularly interact with. This is why strategic engagement — thoughtfully replying to accounts in your niche — is so powerful. You're essentially training the algorithm to surface your content to their audience.

?Frequently Asked Questions

No — the algorithm creates a personalised feed for each user based on their engagement history, network graph, and content preferences. Two users following the same accounts will see different timelines.

Initial ranking happens within seconds of posting. However, the algorithm re-evaluates tweets based on early engagement, so the first 30-60 minutes after posting are critical for performance.

Posting too frequently (10+ tweets/day) can dilute engagement per tweet, which hurts your overall account score. The algorithm rewards accounts whose tweets consistently perform well, not accounts that post the most.

Enjoyed this article? Share it on X.

Share on X
Free Guide

Get the full growth playbook

All the strategies from this article (and more) — compiled into one actionable PDF guide.

  • Step-by-step growth framework
  • Real tweet templates included
  • Algorithm cheat sheet

No spam. Unsubscribe anytime.

S
M
A
R
5,200+ downloads

"This guide alone was worth more than most paid courses."

— Sarah C., Founder

Free weekly newsletter

Level up your X game every week

Join 5,000+ founders and creators getting one actionable growth strategy delivered every Thursday.

No spam everUnsubscribe anytime