When X open-sourced their recommendation algorithm, most people skimmed the announcement and moved on. That was a mistake. I've been building XPilot, a tool that automates X growth for founders, which means I've spent a lot of time inside this codebase trying to understand what the platform actually rewards. And the code tells you, with unusual clarity, what's going on under the hood. For founders trying to build an organic presence, it changes the calculus on almost every content decision. This is what I found: how the algorithm works in 2026, what it rewards, what it suppresses, and what that means for anyone trying to grow a real audience without spending money on promotion. How X Decides What to Show At the highest level, X's recommendation system is trying to answer one question for every piece of content: will showing this to a given user produce a positive outcome? That sounds obvious. But "positive outcome" is more specific than most people realize. The algorithm isn't just looking at whether you engage with a post. It's looking at the type of engagement, how quickly it happened, and what signals that engagement sends about you and the post. type The system uses a machine learning model that predicts expected engagement weighted by engagement type. Not all engagement is equal. Here's roughly how the signal weights stack up, from highest to lowest: Long replies (substantial replies, not just "great point")
Retweets with comment
Profile clicks following exposure to a post
Likes
Regular retweets
Link clicks (weighted lower because they take users off-platform) Long replies (substantial replies, not just "great point") Long replies Retweets with comment Retweets with comment Profile clicks following exposure to a post Profile clicks Likes Likes Regular retweets Regular retweets Link clicks (weighted lower because they take users off-platform) Link clicks The implication is significant: a post that generates a genuine conversation is worth considerably more than a post that racks up likes. If you're optimizing for engagement, you should be thinking about what makes people want to respond and push back, not what makes them passively approve. The Distribution Window One of the most important and underappreciated mechanics is how brief a post's distribution window actually is. When you post something, the algorithm runs a series of initial tests. It shows the post to a small sample of your followers and some non-followers with a likely interest profile. It watches what happens in that window, roughly the first two to four hours. two to four hours If the post generates meaningful engagement during that window, it gets distributed more broadly. If it doesn't, the algorithmic amplification essentially turns off. The post still exists and can accumulate impressions from people who visit your profile directly, but the active promotion stops. This has a major practical implication: when you post matters as much as what you post. when you post matters as much as what you post. Posting when your audience is asleep or distracted means the initial test fails through bad timing, not bad content. The algorithm doesn't give you a second chance. It won't push a post from Tuesday at 3 am to your audience on Thursday because it looks good in retrospect. For most founder audiences (other founders, investors, product people, tech workers), peak activity is roughly 8 am to 11 am Eastern Time on weekdays, with a secondary window around 12 pm to 2 pm. 8 am to 11 am Eastern Time on weekdays Follower Quality vs. Follower Count Here's the part that surprises most people when they first understand it properly. The algorithm doesn't just look at your follower count when deciding how widely to distribute your posts. It looks at the engagement rate of your existing followers. Specifically, it considers how frequently your followers engage with your content as a signal of account quality. engagement rate of your existing followers What this means in practice: an account with 1,000 followers where 200 of them regularly reply and engage is treated very differently from an account with 10,000 followers where almost nobody engages. The first account has a higher "trust score" because the people who chose to follow it actually care about what it says. The algorithm infers it's worth showing to new people. The second account, despite having 10x the followers, may get less favorable initial distribution because the engagement rate signal is telling the algorithm that most of those followers aren't that interested. This is why chasing follower count through follow-back strategies, buying followers, or viral posts that attract people with no real interest in your work can actually hurt your distribution in the medium term. You're inflating a metric while degrading the underlying signal. Twenty followers who reply to most of your posts are worth more algorithmically than two hundred followers who scroll past everything. What Gets Suppressed Knowing what the algorithm rewards is useful. Knowing what it actively suppresses is essential. External links in the main post body. Posts that include links get significantly less distribution than link-free posts. This is one of the clearest patterns in the algorithm data, and it's been consistent for over a year. If you need to share a link, putting it in the first reply rather than the post itself generates meaningfully more impressions. External links in the main post body. Low-effort engagement bait. "Like if you agree," "RT if you're a founder," and similar formats are detected and penalized. The algorithm has become quite good at identifying posts that are structured to harvest engagement without generating a genuine signal. These posts often see an initial spike followed by rapid suppression. Low-effort engagement bait. Irregular posting cadence. Accounts that post sporadically are treated differently from accounts with a consistent cadence. If you disappear for three weeks and come back, you don't just pick up where you left off. The algorithm needs to re-evaluate your account, and in the meantime, your posts are shown more conservatively. Irregular posting cadence. Low-quality follower ratios. If a large percentage of your followers are inactive, bot-like, or have very low engagement records themselves, this degrades the algorithm's view of your account. It treats follower quality as a proxy for your content quality. Low-quality follower ratios. Excessive same-format content. Posting the same type of content repeatedly, at the same length, in the same structure, correlates with lower distribution over time. Variation in format signals a more natural posting pattern and keeps engagement rates healthier. Excessive same-format content. The Reply Strategy Replies are the highest-value engagement type in the algorithm, and they're also one of the most underused growth levers for founder accounts. When you leave a substantive reply on a high-traffic post in your space, a few things happen. Your reply is visible to everyone who views that post. If it generates its own engagement, it gets more exposure. People who find it interesting click through to your profile. This is one of the most effective ways to get discovered by people who are already interested in your topic area. You're hitching a distribution ride on someone else's post by adding genuine value to the conversation. The keyword is genuine. A reply that just says "great thread" does nothing for you. A reply that takes a clear position, adds a counterpoint, or asks a sharp question that moves the conversation forward, that gets engagement of its own and drives profile visits. Spending 15 minutes each morning, leaving three or four substantive replies in your space, will compound into significant audience growth over time. It's one of the highest-leverage activities on the platform for accounts that don't yet have large followings. Thread Strategy in 2026 Threads remain one of the highest-performing content formats for founder accounts, but the mechanics have shifted. The old approach of posting "1/" as a single tweet and replying to yourself still works, but native thread formatting (using X's built-in thread composer to write the whole thing before posting) performs modestly better. The algorithm appears to treat native threads as a single content unit and evaluates aggregate engagement more favorably. Length matters. The threads that perform best are long enough to deliver genuine value but not so long that completion rate drops significantly. In practice, five to ten posts is the sweet spot. Longer than that and you're asking for a commitment most readers won't make unless you're already well-established. five to ten posts is the sweet spot The opening post is the most important. It needs to earn the click to read more. The best thread openers are specific, counterintuitive, or promise a clearly useful payoff. "How I went from 0 to 5,000 followers in 6 months" is a better opener than "A thread on X growth." Not because one is clickbait and the other isn't, but because one gives you a specific, concrete thing to evaluate and the other doesn't. Timing and Scheduling Given everything above, timing deserves more attention than most founders give it. For a typical founder audience, the data points consistently toward early mornings on weekdays, Tuesday through Thursday specifically. Monday often sees lower engagement because people are catching up from the weekend. Friday afternoon drops sharply. Weekends are inconsistent. early mornings on weekdays, Tuesday through Thursday The morning window matters for two reasons. First, it's when your audience is active and will generate the early engagement the algorithm needs to see. Second, it gives the post a full workday to accumulate impressions before the audience goes offline. If you're posting manually, pick a consistent time in your audience's morning and stick to it. Consistency in timing trains your audience to expect content from you, which increases the likelihood of early engagement. If you're automating scheduling, use your actual analytics data to find when your specific followers are most active. The platform shows this in the analytics dashboard, and it varies enough by account that general rules of thumb are only a starting point. What This All Means for Your Content Strategy Putting it together, here's what a founder should actually change: Optimize for replies over likes. Write posts that invite a response. Ask genuine questions. Take positions that not everyone will agree with. A post that generates 10 substantive replies is algorithmically more valuable than a post that generates 100 likes. Optimize for replies over likes. Build a genuinely engaged core before scaling. Don't chase follower count. Spend your early months building relationships with 50 or 100 people who actually care about your work. Those relationships create the engagement floor that makes the algorithm treat your account favorably as you scale. Build a genuinely engaged core before scaling. Post every day, at the same time. Consistency and timing are the least exciting parts of X strategy, but they have an outsized effect on distribution. A mediocre post at the right time with an engaged audience will outperform a great post at the wrong time with a passive one. Post every day, at the same time. Keep links out of the post body. Put links in the first reply, not in the post itself. It's a small adjustment that consistently improves reach. Keep links out of the post body. Use replies as a discovery tool. Every day, find two or three high-traffic posts in your space and leave a reply that's worth reading on its own. Over months, this compounds into meaningful profile traffic. Use replies as a discovery tool. Vary your format. Mix short observations, longer threads, questions, and direct opinions. Format variety keeps your engagement rate healthy, which keeps your distribution healthy. Vary your format. The One Thing Most People Miss The algorithm is designed by people who want the platform to be worth using. At every level, it's trying to surface content that people genuinely want to see. This means the best X strategy is not really a strategy. It's just worth following. Sharing things that are interesting, specific, and honest. Engaging with your audience like they're people, not just a number. The founders who grow the fastest on X are not the ones who've studied the algorithm most carefully. They're the ones who have the most interesting things to say and the most genuine way of saying them. Understanding the algorithm helps you avoid mistakes and time your posts better. But it doesn't give you the thing that actually drives growth, which is a reason for people to care what you think. That part, you have to earn. I'm building XPilot, an AI-powered X growth autopilot for founders, which is why I went deep on the algorithm code in the first place. Take that context into account. I'm building XPilot, an AI-powered X growth autopilot for founders, which is why I went deep on the algorithm code in the first place. Take that context into account. XPilot

This story contains new, firsthand information uncovered by the writer.

This story contains AI-generated text. The author has used AI either for research, to generate outlines, or write the text itself. 

I Read X's Open-Source Algorithm: Here's What Actually Matters in 2026

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Building the "Spotify Wrapped" of Product Hunt: A 48-Hour Technical Sprint to Map the Maker Economy

I'm Building a Tower Defense Game Where VCs Are the Enemy

Preparing Your Phone for a Protest: How to Protect Yourself

How to Future-Proof Your Technical Documentation for AI and RAG Systems

3 Free Marketing Tactics I Used to Grow 'Who Raised?' To $2K MRR

I Made a Productized Service for the Solopreneur GTM and the Build-in-Public Crowd: Here's How

Building the "Spotify Wrapped" of Product Hunt: A 48-Hour Technical Sprint to Map the Maker Economy

I'm Building a Tower Defense Game Where VCs Are the Enemy

Preparing Your Phone for a Protest: How to Protect Yourself

How to Future-Proof Your Technical Documentation for AI and RAG Systems

3 Free Marketing Tactics I Used to Grow 'Who Raised?' To $2K MRR

I Made a Productized Service for the Solopreneur GTM and the Build-in-Public Crowd: Here's How

Light-Mode

Classic

Newspaper

Minty

Dark-Mode

Neon Noir

Minty

HN StartUps