← blog

The X (Twitter) algorithm in 2026: what decides your reach?

What X's main-feed ranking looks at — why the first minutes matter, the signals it rewards and penalizes, and how to write your tweet accordingly.

algorithm · June 17, 2026

What ranks the feed?

X's "For You" feed isn't chronological — an estimated engagement probability is calculated for each tweet and it's ranked accordingly. The core signals the model looks at: the replies/retweets/likes a tweet gets in its first minutes, time spent on the content, and especially replies. A reply weighs far more than a like because it's a genuine conversation signal.

  • +The first 30 minutes are critical: early engagement determines whether the tweet is shown to a wider audience.
  • +Reply > retweet > like: that's the weighting order. Content that pushes people to respond wins.
  • +Time spent on the profile and opening/reading the tweet are positive signals too.

Penalized patterns

The algorithm penalizes some patterns by lowering reach. Most of these are in the "engagement bait" class:

  • +"RT this", "like and follow", "👇 thread" and other overt engagement begging.
  • +Many hashtags per tweet (1–2 is enough; 3+ is a spam signal).
  • +External links: X wants you to stay on its platform, so a tweet with a link is somewhat suppressed — putting the link in the first reply is a common workaround.
notetweetloom's generation engine filters these patterns deterministically at generation time: hashtag spam, bait phrases, and cliché openings are penalized in the 0–100 score, and content below the threshold isn't posted in autonomous mode.

How to write in practice

  • +The first line is the hook: open with a specific claim, number, or tension — avoid generic sentences that stick to every day like "AI is everywhere."
  • +Invite a reply, but not with a tacked-on generic question — create a genuine difference of opinion that arises from the content.
  • +One clear idea: instead of cramming three ideas into 280 characters, say one idea sharply.
For questions, info@tweetloom.com