How Does the X (Twitter) Algorithm Work in 2026?

The X (formerly Twitter) algorithm is the recommendation system that determines which tweets appear in each user's "For You" feed, ranking content based on engagement signals, network relationships, content type, and relevance scoring. X open-sourced a significant portion of its recommendation algorithm in March 2023 on GitHub, giving the public an unprecedented look at how a major social platform ranks content. With an estimated 500-600 million monthly active users, understanding this algorithm is essential for anyone using X for business growth.

For You vs Following: Two Different Feeds

X offers two primary feed experiences:

For You (algorithmic). This is the default feed that most users see. It uses X's recommendation algorithm to surface a mix of tweets from accounts you follow, accounts the algorithm thinks you will find interesting, and trending content. The For You feed is where algorithmic reach happens - it is how your tweets reach people who do not follow you.

Following (chronological). This feed shows only tweets from accounts you follow, in reverse chronological order. There is no algorithmic ranking. Tweets appear purely based on when they were posted.

For content creators and brands, the For You feed is where growth happens. The Following feed provides consistent visibility to existing followers, but the For You feed is the discovery engine that puts your content in front of new audiences.

How the Algorithm Ranks Tweets

Based on the open-source code and analysis by researchers, the X algorithm uses a multi-stage process:

Stage 1: Candidate Generation

The algorithm first assembles a pool of roughly 1,500 candidate tweets to potentially show in your feed. It pulls from three main sources:

• In-network tweets - Recent tweets from accounts you follow (roughly 50% of the For You feed)
• Out-of-network tweets - Tweets from accounts you do not follow that the algorithm predicts you will like, based on what similar users engage with (roughly 50% of the feed)
• Trending and breaking content - Tweets gaining rapid engagement that are relevant to your interests

Stage 2: Ranking With Machine Learning

Each candidate tweet gets scored by a roughly 48-million-parameter neural network that predicts how likely you are to engage with it across multiple dimensions (like, reply, retweet, bookmark, profile click, dwell time). These predictions are combined into a single ranking score. The key signals include:

Engagement velocity. How quickly a tweet accumulates engagement after being posted. A tweet that gets 10 replies in the first 15 minutes signals higher quality than one that gets 10 likes over several hours.

Reply weight. The open-source code revealed that replies are the baseline engagement signal (approximately 1x weight). They are the strongest public signal because threaded replies and conversations boost a tweet significantly.

Retweet and quote tweet weight. Retweets count for roughly 1x the base engagement score. Quote tweets, which add commentary, are weighted similarly but also create a new tweet that can accumulate its own engagement.

Bookmark weight. Bookmarks (saves) carry approximately 1x weight - surprisingly high given that they are private and not visible to others. The algorithm treats bookmarks as a strong "high-intent" relevance indicator, signaling content worth returning to.

Like (favorite) weight. Likes carry approximately 0.5x the base weight - lower than replies, retweets, and bookmarks. Passive likes matter less than active engagement.

Negative signals are devastating. The open-source code showed that negative actions - "Not interested," blocks, mutes, and reports - carry approximately -74x weight. A single negative action can undo dozens of positive engagements.

Profile clicks and follows. If a tweet leads users to click on your profile or follow you, that is a strong signal of high-quality content.

Dwell time. How long users spend looking at a tweet. If people stop scrolling and read the full tweet, it signals interest even without an explicit engagement action.

Stage 3: Filtering and Diversification

After scoring, the algorithm applies filters:

• Author diversity - Your feed will not show too many tweets from any single account in a row
• Content type diversity - The algorithm mixes text, image, video, and link tweets
• Negative signals - Tweets that get hidden, muted, or reported receive negative scoring
• Freshness - More recent tweets get a time-decay boost, though exceptional older tweets can still appear

Key Algorithm Factors to Know

X Premium Boost

X Premium (the paid subscription, formerly Twitter Blue) provides a measurable boost in the For You feed. The open-source code confirmed that verification status acts as a positive ranking signal. Premium subscribers also get:

• Higher character limits for posts (up to 25,000 characters)
• Ability to edit tweets within a time window
• Higher priority replies in conversations
• Revenue sharing eligibility
• Access to X's AI assistant (Grok)

Whether the Premium boost is worth $8-16/month depends on your goals. For active content creators, the algorithmic boost alone often justifies the cost.

External Links Get Reduced Reach

Similar to other social platforms, X reduces the algorithmic reach of tweets containing external links. The platform prefers to keep users on-platform. If you need to share a link, common strategies include:

• Put the link in a reply to your tweet rather than the main tweet
• Use a thread format where the first tweet has the hook and a later tweet has the link
• Post the valuable content natively on X and mention the link verbally

Negative Engagement Signals

The algorithm also tracks negative signals that reduce your tweet's distribution:

• Users hiding your tweet
• Unfollows triggered by your tweet
• Users marking your content as "Not interested"
• Reports or blocks

Consistently creating content that triggers negative signals will reduce your overall account reach over time.

How Content Type Affects Reach

Different content formats get different levels of algorithmic support:

Text tweets remain the bread and butter of X. Clear, concise text with a strong opening line performs consistently well.

Image tweets receive an algorithmic boost. The algorithm can analyze image content and tends to favor original images over stock photos.

Video tweets get a distribution boost, particularly native video uploaded directly to X rather than linked from YouTube or other platforms. Short videos under 60 seconds tend to perform best.

Threads (multi-tweet posts) signal effort and depth. Long threads that hold attention through multiple posts often outperform single tweets on the same topic.

Polls generate high engagement because voting is a low-friction interaction. However, poll engagement may not translate to follower growth as effectively as other formats.

Understanding the X algorithm comes down to one principle: create content that sparks genuine conversation. The algorithm heavily weights replies and extended engagement over passive interactions. If your content makes people want to respond with their own thoughts, the algorithm will distribute it to a wider audience.

For a practical step-by-step guide on applying these algorithm insights, see how to grow on X from zero followers.