[SNS] X Algorithm Hacks: Post Smart, Grow Steadily

Introduction

Hello, I'm Easygoing.

In my previous post, I introduced how note and X are excellent platforms for sharing information.

This time, let’s dive into X’s timeline display algorithm.

Elon Musk is Open to Technology Transparency

Since taking over as X’s leader, Elon Musk has been open about sharing technical details.

In March 2022, while planning to acquire Twitter (now X), Musk conducted a poll asking, “Should Twitter’s algorithm be open-sourced?”

With over 1 million votes, 82.7% supported transparency. Following Musk’s acquisition of Twitter, X’s timeline algorithm was made public on March 31, 2023.

• X Official Blog (March 31, 2023)
• X Algorithm (GitHub)
• X AI Algorithm (GitHub)

Today, we’ll explore this publicly released algorithm.

X’s Timeline Display Process in Three Stages

Posts displayed on X’s timeline are determined through the following steps:

flowchart TB

    A1(Relevant Users)

subgraph Post Selection
    B1(750 posts from followers)
    B2(750 posts from non-followers)
end

subgraph Ranking
    C1(Analyze past behavior)
    C2{Heavy Ranker<br>AI predicting<br>user behavior}
    C3(Calculate score with weights)
end

subgraph Post-Processing
    D1(Exclude blocked/muted posts<br>Exclude duplicate posters<br>Add advertisements)
end

E1(User Timeline)

A1-.->B2
B1-->C1
B2-->C1
C1-->C2
C2-->C3
C3-->D1
D1-->E1

1. Selecting 1,500 Posts

X first selects 1,500 posts to display on a user’s timeline.

These posts come from users relevant to you: 750 posts from users you follow and 750 posts from relevant non-followed users.

Who Are Relevant Non-Followed Users?

Relevant non-followed users meet these criteria:

• Belong to the same cluster (group)
• Share common followers/following
• Engage with the same posts (likes or reposts)

X’s clusters are automatically updated every three weeks based on user activity.

For example, if you frequently post about POP music, the algorithm places you in the POP music cluster and recommends POP music-related posts to your timeline.

2. Ranking Posts

Next, X ranks the 1,500 selected posts to determine their order on the timeline.

Posts likely to generate user interactions (likes, reposts, replies) appear higher.

2-1. Analyzing Past Data

X predicts whether users will interact with a post with over 87% accuracy based on past data analysis.

This analysis primarily focuses on viewers but also considers poster-related factors:

• Activity level (total days with at least one interaction)
• User reputation

As a poster, liking at least once daily and avoiding actions that harm your reputation can boost your post’s ranking.

2-2. AI Optimization

X uses an AI called Heavy Ranker to predict user behavior with even higher accuracy.

Heavy Ranker predicts specific user actions when viewing a post.

2-3. Scoring and Ranking

Heavy Ranker assigns probabilities, which are multiplied by specific weights to calculate a post’s score.

Higher weights are given to actions that sustain user engagement, like replies or profile views. Conversely, posts likely to receive negative reactions (mute, block, report) get large negative scores, lowering their ranking.

Does X Fuel Controversy?

SNS algorithms, including X’s, assign strong weights to replies to encourage prolonged app usage.

This means that when conflicting opinions lead to a reply chain, the timeline gets flooded with those replies.

Negative actions like muting or reporting slightly lower a post’s score, but multiple replies can outweigh this penalty, boosting the score. Since X’s algorithm continues displaying conflicting opinions after replies, ignoring opposing views early on is key.

3. Post-Processing

After ranking, X performs post-processing:

• Excludes posts from blocked or muted users
• Removes duplicate posts from the same user
• Adds advertisements

This finalizes the posts displayed on your timeline.

Follower-Following Ratio (FF Ratio) Up to 0.6

The 2023 code includes an account credibility adjustment based on the follower-following ratio.

For accounts following over 500 users, a follower-following ratio above 0.6 incurs a credibility penalty.

Accounts with significantly more followings than followers are flagged as spam accounts, while those with more followers than followings are deemed more trustworthy. Abruptly unfollowing users may also trigger penalties, so maintaining a balanced ratio is advisable.

Other Algorithms Beyond the Public Code

In addition to the published code, X implements the following algorithms:

Verified Accounts Are Prioritized

X’s blue verification badge, available through a premium subscription (¥900/month), boosts post visibility.

External Links Lower Visibility

Elon Musk advises including external links in replies rather than main posts.

This suggests X’s algorithm downranks posts with external links, as they may lead to user departure from the platform. While promoting articles via SNS is common, posts with external links are less likely to appear on X.

Bonus: Should You Post Images?

A code section published on March 31, 2023, was retracted five days later as “not actually used”.

It included weights favoring posts with images or videos and penalizing posts with multiple hashtags.

While it’s unclear if these weights are currently applied, they indicate the algorithm’s direction, so it’s worth considering when possible.

Conclusion: Keep X Casual!

• Like at least once a day
• Use external links sparingly
• Ignore opposing opinions

Many claim to have “hacks to skyrocket X impressions,” but few are backed by evidence.

This post referenced the code published in March 2023, but Elon Musk has stated that X’s algorithm changes every 24–48 hours, so it may already differ.

The current algorithm’s details are unknown, and obsessing over rankings can make SNS less enjoyable. I plan to keep using X casually, avoiding unnecessary controversies.

Thank you for reading!