The X Algorithm Explained: Distribution vs Engagement

Understanding how X decides what to show people is the difference between reaching thousands and reaching dozens. Most advice focuses on content quality, which matters, but ignores the mechanical reality of how distribution actually works.

Distribution and engagement are different problems with different solutions. Distribution is about who sees your content. Engagement is about who interacts with it. The algorithm treats them differently, and so should you.

How Distribution Works

Every day, X processes around 500 million tweets. The algorithm can't show all of them to everyone, so it filters ruthlessly through a three-stage pipeline.

The system uses a two-stage pipeline. First, a retrieval stage (using a Two-Tower model) sources candidate tweets that might be relevant to each user based on who they follow, what topics they engage with, and what's trending. Second, a Grok-based transformer model called Phoenix ranks those candidates based on predicted engagement, with final filtering to balance content types and freshness.

The "For You" feed splits roughly 50/50 between in-network content (from accounts you follow) and out-of-network content (from accounts you don't). An out-of-network (OON) scorer adjusts scores for content from accounts you don't follow, while an author diversity scorer prevents any single account from dominating your feed.

The Weights That Matter

X's algorithm assigns different values to different engagement types. Understanding which signals matter explains why some posts reach thousands while others disappear.

The confirmed positive engagement signals include: favorites (likes), replies, retweets, quote tweets, shares (including via DM and copy link), photo expansions, clicks, profile clicks, video quality views (vqv), dwell score, dwell time, and follows of the author. The algorithm weighs these signals heavily, with replies, retweets, and quote tweets carrying significantly more weight than simple likes.

The exact numerical weights are not publicly disclosed in the current algorithm (X has marked these as excluded for security reasons), but the hierarchy is clear: conversation-driving actions like replies and retweets are valued far more than passive engagement like likes.

Negative signals are equally important:

• Reports: Catastrophic impact on distribution
• Blocks or mutes of the author: Severe negative weight
• "Not interested" clicks: Significant negative signal

One spam report can tank a post's distribution entirely. The algorithm is designed to be extremely cautious about content that generates negative reactions.

The Premium Divide

This is the uncomfortable reality of X in 2025-2026: Premium subscribers get dramatically better distribution.

Buffer's analysis of 18 million+ posts found Premium accounts receiving approximately 10x the impressions of free accounts for identical content. Premium+ subscribers receive an even larger boost. Since March 2026, non-Premium accounts posting links receive effectively zero engagement.

Premium accounts get a 4x boost for in-network content and a 2x boost for out-of-network content. Their replies appear higher in threads by default. They're eligible for the creator program and monetization.

For serious growth, Premium has become effectively mandatory. The alternative is working four times as hard for the same results.

What Controls Distribution

Several factors determine who sees your content before anyone has a chance to engage with it:

Account status: Premium vs free is the biggest single factor. Verified accounts with established engagement history get preferential treatment.

Content type: Native content beats external links. The algorithm penalises links,removing them from a post increases views by an average of 270%. Video content receives 2x organic reach versus static images.

Timing: Early engagement velocity matters enormously. A post getting 50 engagements in the first hour dramatically outperforms the same 50 engagements over 24 hours. The algorithm tests with small audiences first, then expands distribution based on performance.

Account reputation: X uses various signals to assess account quality, including follower-following ratio, posting consistency, and engagement patterns. Accounts with poor reputation signals get reduced distribution on all content.

What Controls Engagement

Distribution gets your content in front of people. Engagement determines whether they do anything with it.

The hook: The first line is everything. Users decide in about a second whether to keep reading. A weak opening kills engagement before it starts.

Dwell time: The algorithm tracks how long people look at your content. Less than 3 seconds of attention signals low quality. Content that keeps attention for 15+ seconds on the detail view gets boosted.

Conversation depth: Reply chains generate the highest algorithmic value. A tweet that sparks 10 back-and-forth conversations outperforms one with 50 isolated likes. Learn how to craft replies that spark conversation.

Format match: Video, images, GIFs, and text perform differently for different audiences. What works for your niche depends on what your specific audience engages with.

The Recent Changes

The platform has evolved significantly in the past year:

Grok AI integration: As of November 2025, even the Following feed is no longer chronological by default. Grok AI ranks posts from accounts you follow based on predicted relevance and engagement. Users can access chronological feeds through settings, but most don't.

Small account boost: X now actively promotes content from emerging creators, though the effect is modest compared to established account advantages.

Link penalties persist: Despite announcements about removing link suppression, testing shows external links still dramatically underperform native content. The workaround,posting links in replies,remains necessary.

The Practical Takeaway

Distribution and engagement compound each other. Good engagement signals teach the algorithm to give you better distribution next time. Better distribution means more chances to generate engagement.

The strategy that follows from this:

For distribution, optimize your account status (Premium if possible), content type (native over links), and timing (post when your audience is active, stay engaged in the first 30 minutes).

For engagement, optimize your hooks, create content that rewards attention, and design for conversation rather than passive consumption.

Neither alone is sufficient. Great content with zero distribution reaches no one. Wide distribution of weak content generates no engagement, which trains the algorithm to stop showing your content.

The algorithm isn't trying to hurt you; it's trying to predict what will keep users engaged with the platform. When you create content that genuinely keeps attention and sparks conversation, you're aligned with what the algorithm rewards. Understanding this difference between virality and sustainable growth is essential.