The ranking signals in X's algorithm, and what each one is worth

The ranker doesn't score your post on a single "engagement" number. It predicts the probability of each action a viewer might take, multiplies each by that action's weight, and sums them. Understanding those weights tells you exactly what to optimize for.
The positive signals (earn these)
- Copy-link shares (+20) - someone copies your link and pastes it somewhere else. The heaviest positive weight in the table.
- Replies (+5, or +20 from a mutual follow) - conversation is worth far more than a passive like, and 4× more again between accounts that follow each other back.
- DM shares and quote posts (+5 each) - forwarding in private counts as much as quoting in public.
- New follows (+4) - the post converted a viewer into a follower.
- Reposts (+1) - the post is worth spreading to someone else's audience.
- Continuous read time (+0.004 per second, unbounded) - and only that. The threshold dwell head and the profile-click head both ship at 0.
- Likes (+0.5) - positive, but a small fraction of a reply's value.
The negative signals (avoid these)
- Reports (-234), mutes and blocks (-90), "not interested" (-43.2) - large negative weights that can sink a post fast.
- Spam and low-quality classifiers - patterns that look automated or engagement-baiting.
- Muted keywords - posts matching a viewer's muted words are removed outright, before scoring even starts.
- Hashtag stuffing - three or more reads as spam to the classifier that screens posts before ranking.
One thing that is not on that list: external links. The published ranker carries no link penalty - a link open scores +0.2. What a link costs you is the read time that ends the moment someone clicks away.
Probability times weight is the whole game
A draft that slightly raises the chance of a heavily-weighted action (a reply-back) beats one that chases a lightly-weighted one (a like). This is why "write to start a conversation" consistently outperforms "write to get likes" - the math literally favors it.
How to use this when you write
Before you post, ask: does this invite a reply? Is it worth stopping to read to the end? Would someone repost it to their own audience? If yes, you are pointed at the heavy weights. Scoring a draft against the real signals turns that gut check into a number you can improve.
X Radar drafts every post in your cloned voice and scores it against these exact signals in under a second, before you ever hit post. Sign in with X to try it free.
Written by
I built X Radar by reading X's open-source ranking code line by line. Every signal it scores traces to a named file in xai-org/x-algorithm at a pinned commit - so the numbers here are checkable, not folklore.
Keep reading
How the X (Twitter) algorithm actually works in 2026: the signals that decide your reach
A plain-English breakdown of X's open-source ranking algorithm in 2026 - the signals that lift or bury a post, and how to write for them.
6 min readX monetization requirements in 2026: what it actually takes to get paid
The real eligibility bar for X's ad revenue sharing in 2026 - Premium, 500 followers, 5M impressions - and why the algorithm decides your payout.
6 min readWhy replies beat posts: the reply-back multiplier explained
The highest-value move on X isn't posting - it's an early, sharp reply the author replies back to. Here's why the algorithm rewards it so heavily.
5 min read
Stop guessing what the algorithm wants.
X Radar drafts in your voice and scores every post against X's real ranking signals - before you publish.
Sign in with X