How to Change the Twitter Algorithm (Your Feed + Reach)

Two Different Problems, One Keyword

When people search "how to change the Twitter algorithm," they're usually asking one of two very different questions. Either they want to change what they see - clean up a cluttered feed, stop seeing content they don't care about, and train X to show them useful stuff. Or they want to change how the algorithm treats their own posts - get more reach, break out of a low-impression plateau, and actually grow.

Both are solvable. But they require completely different approaches. I'll cover both here - plus a section on the technical mechanics underneath all of it, because if you understand what the system is actually doing, every tactic makes more sense.

How the X Algorithm Actually Works (The Technical Reality)

Before you can change the algorithm, you need to know what you're dealing with. X is one of the only major social platforms that has open-sourced its recommendation code, which means we have more hard data on this than almost any other platform.

Every time you open the For You timeline, the system executes a three-stage pipeline. First, it pulls roughly 1,500 candidate posts from an enormous pool - about 50% from accounts you follow, 50% from accounts you don't. It narrows those down from 500 million daily posts in under 200 milliseconds. Second, a neural network called the Heavy Ranker scores each candidate post based on your predicted likelihood to engage with it. Third, a filtering stage removes content from muted or blocked accounts, applies content moderation, enforces diversity rules so a single author doesn't dominate your feed, and blends in ads, Spaces, and trending topics.

The Heavy Ranker doesn't just count likes. It runs a transformer model - now Grok-powered - that analyzes thousands of features per post and predicts how you'll interact across multiple dimensions: whether you'll like it, reply to it, repost it, bookmark it, click through to the author's profile, or spend meaningful time reading it. Each post gets a final score equal to the probability of each action multiplied by that action's weighted importance factor.

The Engagement Weight Hierarchy

Here's the part most people get completely wrong. The engagement weights from the open-sourced algorithm code tell a very specific story about what the system actually values. The simplified scoring formula from the published code looks like this:

Likes x 0.5 + Retweets x 1.0 + Replies x 13.5 + Profile Clicks x 12 + Bookmarks x 10 + Link Clicks x 11

But there's a layer on top of this that most guides miss. When you reply to a post and the author replies back - creating an actual conversation thread - that interaction carries a weight of roughly +75. A two-way conversation between author and commenter is weighted at approximately 150 times the value of a single like. The algorithm doesn't just want replies. It wants conversations with depth.

What this means practically: a post with 50 thoughtful replies that the author engaged with will crush a post with 500 passive likes in algorithmic distribution. Most people have been optimizing for the lowest-value signal on the board.

Negative signals are even more dramatic. The open-source code shows that negative actions - "Not Interested," blocks, mutes, reports - carry approximately -74x weight. A single user reporting a post can undo dozens of positive engagements. Your account reputation score (called TweepCred internally) takes a hit every time someone mutes or blocks you, and once your TweepCred drops below a certain threshold, only a fraction of your posts are eligible for distribution at all.

SimClusters: Why Topical Consistency Matters More Than Follower Count

The out-of-network half of your feed - the 50% of posts from accounts you don't follow - is powered primarily by a system called SimClusters. The algorithm groups users into roughly 145,000 overlapping topic clusters based on shared engagement history and follow patterns. When you post, the Grok model reads the semantic meaning of your content (not just keywords) and matches it to the most relevant clusters. Posts that resonate with users in those clusters get tested against wider audiences in a feedback loop.

Here's what this means for growth: your follower count matters less than your topical relevance. A 500-follower account posting consistently high-quality content in a specific niche can land in the For You feed of users with hundreds of thousands of followers, if the content matches the right cluster. But if you normally post about B2B sales and suddenly start tweeting about football, the algorithm loses its map of who should see your content. Your existing audience doesn't engage with the off-topic post, the algorithm reads that as a quality failure, and your distribution gets suppressed - even on your next on-topic post.

The practical implication: pick two or three topics you can own and post about them consistently. Not because "niching down" sounds good in theory, but because the algorithm literally cannot find your audience unless it knows what cluster you belong to.

TweepCred: Your Account-Level Reputation Score

Every X account carries a TweepCred score from 0 to 100, calculated using a weighted PageRank approach. Factors include account age, follower-to-following ratio, engagement quality (who engages with you matters more than raw numbers), and interaction patterns with high-quality users. Accounts that follow far more people than follow them back get penalized. Getting muted, blocked, or reported directly tanks this score. And accounts with very low TweepCred scores have their content distribution throttled regardless of how good a given post is.

There's no shortcut to building TweepCred. It compounds over time through consistent quality posting, genuine engagement, and avoiding the behaviors - follow-for-follow schemes, spam engagement, buying followers - that trigger the system's credibility penalties.

Part 1: How to Change What You See in Your Feed

The X algorithm curates your "For You" feed using a machine learning model that tracks your behavior - your likes, replies, reposts, profile visits, and even how long you pause on a post. Every action you take teaches the system what to show you more or less of. If your feed is a mess, it's because you've been accidentally training it to show you garbage.

Here's how to reset that.

Switch to the Following Tab (Immediately)

The simplest fix: stop using the "For You" tab as your default. The "Following" tab shows posts from accounts you actually chose to follow, mostly in order of recency. It removes the algorithmic noise. Tap the "Following" label at the top of your home feed to switch. Use this as your primary tab while you fix the rest.

One thing worth knowing: X has been updating the Following tab to incorporate some Grok-based sorting by predicted engagement, though users can still toggle to a fully chronological view. Fewer users actively switch to this feed in the first place, which means most of the platform's distribution still runs through the For You algorithm - but for your own consumption purposes, the Following tab gets you back to basics fast.

Use "Not Interested" Aggressively

On any post in your feed that doesn't belong there, tap the three-dot menu and hit "Not interested in this post." Do this consistently across several posts on the same topic, and X will start getting the message. Pair this with muting keywords - go to Settings - Privacy and Safety - Muted Words and add any terms, handles, or hashtags you're sick of seeing. Muting words permanently filters those topics from your timeline until you reverse it.

You can also now use Grok directly to customize your feed. X has rolled out the ability for users to input natural language commands to tell the algorithm what they want more or less of. Telling Grok "show me less politics" or "more B2B content" gets you faster results than waiting for the system to infer it from your behavior. This is one of the more useful feed control features that most people haven't found yet.

Unfollow Dead Weight

Your follow list is the raw input for your feed. If you followed 800 accounts over five years and half of them are inactive or off-topic, your algorithm is working with bad data. Audit your following list. Unfollow anyone who consistently posts content you scroll past without reading. The algorithm reads your scroll-past behavior as a negative signal anyway - it's just slower than muting.

While you're at it, look at who you're following that you've never actually interacted with. Those accounts contribute noise to the system without contributing signal. Trimming your follow list is one of the fastest feed resets you can do.

Engage Intentionally with What You Want More Of

The most powerful feed-training tool you have is your own engagement. When you reply to, bookmark, or spend time reading a post, X's system registers that as a strong positive signal. Don't just like posts from accounts you want to see more of - reply to them. Replies carry far more algorithmic weight than likes. A week of focused, intentional engagement on the right topics will noticeably shift your feed.

Bookmarks are especially powerful for training your feed. The algorithm interprets bookmarks as a signal that content has lasting value - not just impulse engagement. Bookmark the posts you actually want to read again, and the system learns to surface more content like it.

Clean Up Your Data Profile

X stores inferred ad interests and personalization data based on your activity history. You can review this under Settings - Privacy and Safety - Data Sharing and Personalization. Toggle off personalized ads and review what X thinks you're interested in. This won't nuke your feed overnight, but it stops old data from dragging your recommendations backward.

Bottom line on feed control: the algorithm is a feedback machine. Feed it clear, consistent signals about what you want, and it adjusts. The problem for most people is they've been sending mixed signals for years.

Part 2: How to Change How the Algorithm Treats Your Posts

This is where it gets interesting - and where most Twitter advice is completely wrong. Let's break down the specific levers you can actually pull.

Understand the Scoring Formula (And Stop Optimizing for Likes)

Most creators optimize for the wrong metrics. They focus on likes and retweets while ignoring the signals that actually matter: engagement velocity, dwell time, and negative feedback rate. Based on the open-sourced algorithm code, a like carries approximately 0.5x weight - the lowest positive signal you can get. A reply carries 13.5x. A two-way conversation where the author replies back carries +75. Retweets carry 1x but expand distribution directly. Bookmarks carry 10x.

The practical implication is this: a post that triggers conversation will almost always outperform a post that triggers likes, regardless of the raw number of reactions. Write for responses, not double-taps.

The First 30 Minutes Are Everything

The algorithm applies a steep time-decay factor to posts. A post loses roughly half its potential visibility score every six hours. More critically, the first 30 minutes after posting is the single biggest distribution lever you have. If your post gets a cluster of replies and reposts in the first half-hour, X interprets that as a quality signal and pushes it to a wider audience. If it gets nothing in the first 30 minutes, it's effectively dead.

The strategic implication: post when your audience is actually online. For B2B audiences, data consistently points to Tuesday, Wednesday, and Thursday mornings - specifically the 9 AM to 1 PM window - as the highest-engagement windows. Professionals check X during work hours, and B2B content peaks mid-morning when people are looking for industry updates and insights. If you're targeting a general audience, midday to early afternoon on those same weekdays sees strong overall platform activity.

And if you want to accelerate early engagement, engage with other people's posts right before you publish yours. It warms up your visibility with mutual followers and puts you top of mind. Tools like Taplio can help you schedule posts at optimal times and manage content calendars so you're not winging your posting schedule.

Reply to Your Own Posts - and Reply to Replies

Most people post and walk away. That's a mistake. The algorithm specifically rewards author participation in conversations. When you reply to a comment on your post, the combined interaction carries a dramatically higher weight than the original reply alone. One genuine conversation thread where you engage back is worth more than hundreds of passive likes.

Practically: set a reminder to check your posts 15-20 minutes after publishing and reply to every comment you get in that first window. This isn't just good community building - it's directly generating the highest-weighted engagement signal the algorithm recognizes. A post with 10 replies where you engaged in back-and-forth on each one will score significantly higher than a post with 20 replies you never touched.

Replies Crush Everything Else

Write posts that make people want to respond, not just double-tap and keep scrolling. The most reply-generating post formats are:

• Genuine questions your audience actually has opinions about ("What's the most overrated B2B lead gen channel right now?" generates real answers. "What do you think!" doesn't.)
• Hot takes or mild contrarian positions that invite pushback or agreement
• Incomplete frameworks or lists where people feel compelled to add what's missing
• Posts that start a sentence and invite completion ("The #1 thing I wish I knew before starting my agency was...")
• Direct challenges to conventional wisdom in your niche

The common thread: all of these create a specific gap that people feel compelled to fill. That compulsion is the engine behind reply-generating content.

Bookmarks Are Massively Underrated

Bookmarks carry approximately 10x the algorithmic weight of a like (0.5x). When someone bookmarks your post, it signals to X that the content has lasting value - not just impulse engagement. Content that gets saved is reference material: frameworks, data, how-to breakdowns, specific numbers, step-by-step guides.

If you only post opinions and hot takes, you're leaving bookmark signals on the table. Mix in genuinely useful, save-worthy content and your algorithmic score will reflect it. You can even explicitly prompt bookmarks: "Bookmark this if you want to come back to it" is a legitimate call to action that the algorithm rewards because it generates a high-value signal that's harder to game than a like.

Need ideas for content worth posting? My Daily Ideas Newsletter sends prompts specifically designed to spark the kind of high-engagement content that moves the needle on social.

Kill the External Links (Or Move Them)

X wants to keep users on the platform, and the algorithm actively suppresses posts containing external links. Non-Premium accounts posting external links have seen near-zero median engagement in recent data. This one stings for people who use X to drive traffic, but the data is clear.

The workaround: put your link in the first reply to your main post, not in the post itself. Your main post promotes the content and generates engagement. The reply contains the URL. People who engage with the main post see the link in the replies; your reach on the main post is preserved. This is the standard approach for anyone serious about distribution on X.

Native Media Over Everything

Native video uploaded directly to X significantly outperforms linked YouTube videos or external video. Short-form video under 60 seconds gets the most algorithmic boost. Images stop the scroll and drive engagement. Polls generate easy, high-volume interaction and X specifically rewards them because they signal conversation.

The algorithm scores these formats higher because they generate more dwell time and engagement. A post with an image or video stops the scroll and increases the time a user spends with the content - and dwell time is a meaningful ranking signal. If your current content is 100% text, you're competing with one hand tied behind your back.

Long-Form Posts vs. Threads

X has shifted its algorithm to treat single long-form posts more favorably than multi-tweet threads for distribution purposes. If you have a lot to say, consider using the expanded character limit to say it in one post rather than breaking it into a thread. Threads still work well - and data suggests they get strong engagement when posted during lunch or early evening when people have reading time - but the single long-form post is increasingly the cleaner play for algorithmic reach.

That said, for B2B content and thought leadership, threads remain extremely effective when structured well. The key is: strong hook in post one, each subsequent post in the thread needs to deliver enough value that readers keep scrolling. A thread that drops off in quality after tweet two will have poor completion rates, and the algorithm notices.

Account Reputation Matters (TweepCred)

X assigns every account a reputation score that works like a PageRank for users. It considers follower quality (who follows you matters more than how many), your consistent engagement history, account age, and your follow/follower ratio. Accounts that follow vastly more people than follow them back get penalized. Getting muted, blocked, or reported tanks this score.

The practical moves: don't follow-for-follow. Don't buy followers. Don't use engagement pods that generate fake interactions - the algorithm has gotten better at detecting coordinated inauthentic behavior. Engage meaningfully with high-authority accounts in your niche, because engagement from high-TweepCred accounts amplifies your own reach. And post consistently - accounts with regular posting history score higher than accounts that post sporadically.

X Premium: Worth It for Reach

Premium subscribers receive a 2x to 4x boost in reach compared to non-Premium accounts - this is explicitly coded into the ranking system. Replies from Premium users are algorithmically prioritized to appear at the top of conversation threads. And the penalty for posting external links is substantially lower for Premium accounts. If you're serious about building reach on X as part of a B2B strategy, the math on Premium is fairly straightforward.

The gap between Premium and free account reach has continued to widen. If X is a meaningful channel for your business, treating the subscription as a distribution investment is worth running the numbers on.

Posting Frequency: The Diminishing Returns Problem

One thing the algorithm mechanics make clear: posting volume has diminishing returns, and they kick in faster than most people expect. The algorithm deliberately limits how many posts from a single author appear in any user's feed during a session. Even if you post 10 times a day and all of them are quality content, the algorithm will only show a few to each follower per session.

Data from recent studies suggests accounts posting 3 to 5 times daily achieve the highest median engagement per post. Below that range, you're leaving distribution opportunities on the table - the time-decay factor means a single post only covers one engagement window. But above that range, per-post engagement drops as the algorithm penalizes high volume with low per-post engagement rates.

The ideal approach: post 3 to 5 times per day, spaced at least 2 hours apart. This hits multiple audience activity peaks, gives the algorithm more opportunities to find content that resonates, and avoids the self-cannibalization problem of posting too frequently within a short window. Quality always wins over volume - a pattern of low-engagement posts actively trains the algorithm to expect poor performance from your account.

What Kills Your Reach (The Negative Signals)

Understanding what suppresses reach is just as important as understanding what boosts it. The algorithm is designed to minimize content that users actively dislike, and negative signals are dramatically over-weighted relative to positive ones.

Things that tank your algorithmic distribution:

• Getting muted or blocked: Each one is a direct TweepCred hit. Repeated mutes can put your account in a low-distribution category regardless of content quality.
• "Not Interested" clicks on your posts: These tell the algorithm to reduce distribution of your content to similar users. A cluster of these on a particular content type is a signal to stop posting that type.
• External links in main posts: Actively suppressed. Put them in replies.
• Engagement bait: The algorithm has gotten better at detecting "like if you agree" and similar low-effort engagement tactics. These get suppressed.
• Inconsistent posting: Algorithms favor active accounts. Long gaps between posts reduce your baseline distribution.
• Posting off-topic content: Damages your SimClusters profile and confuses the algorithm's model of who should see your content.

The overarching pattern here is that the algorithm is optimizing for session quality - it wants users to stay on the platform and have positive experiences. Content that causes people to leave, disengage, or actively signal dissatisfaction is exactly what the system is designed to reduce. Build content that makes people want to stay in the conversation, and you're aligned with the algorithm's actual goal.

X Communities, Spaces, and the Explore Tab

Most guides focus entirely on the main feed, but there are other distribution surfaces on X worth understanding if you're trying to maximize reach.

X Communities

Communities are topic-based groups that function as a hybrid of Facebook Groups and Reddit subreddits. Posts in active Communities surface in the For You feed based on topic interest signals - you don't have to be a Community member to see a high-performing post from one. For B2B operators, finding and posting consistently in Communities relevant to your niche gives you a concentrated audience of high-interest users. Engagement from a focused group of highly relevant accounts is one of the fastest ways to build SimCluster authority in a specific topic area.

X Spaces

Spaces are live audio discussions that surface in the For You feed. Real-time content about live Spaces events gets strong algorithmic treatment. If you're comfortable with audio, hosting a regular Space on your topic of expertise is one of the higher-leverage moves available - it combines live engagement (which the algorithm rewards) with community building, and a well-run Space generates downstream social proof when attendees post about it.

The Explore Tab

The Explore tab has been integrated with Grok AI - instead of just a list of trending hashtags, users now often see AI-written summaries of why topics are trending, pulling context from thousands of posts. If your content is consistently high-quality within a trending topic area, there's a path to Explore visibility. This is largely a byproduct of strong For You performance rather than something you can directly engineer, but it's worth knowing the surface exists.

A Week-by-Week Implementation Plan

Reading about the algorithm is one thing. Here's how to actually implement these changes systematically over four weeks.

Week 1 - Feed Reset: Switch your default to the Following tab. Spend 15 minutes aggressively hitting "Not Interested" on content that doesn't belong in your feed. Audit your following list and unfollow inactive or off-topic accounts. Use the Grok feed customization to tell the algorithm what you want to see more and less of.

Week 2 - Engagement Audit: Review your last 30 days of posts in X Analytics. Identify which posts generated the most replies versus the most likes. In almost every case, the reply-generating posts will have outperformed on reach. This is your evidence base for shifting your content approach. Stop posting the types of content that only generated likes.

Week 3 - Content Rebuild: Start posting 3 to 5 times per day using formats specifically designed to generate replies: genuine questions, mild hot takes, incomplete frameworks, contrarian positions. Put any links in replies, not in the main post. Add at least one native image or video per day. Reply to every comment you get within the first 30 minutes of posting.

Week 4 - Systematize: Build a content calendar that covers your two or three core topics. Use a scheduling tool to hit peak posting windows consistently without having to manually time every post. Review your analytics at the end of the week and double down on whatever content type generated the most replies and bookmarks. Remove anything from your rotation that only generates passive engagement.

After four weeks of this, your feed will be dramatically cleaner, your post reach should be measurably higher, and your TweepCred will be trending in the right direction. The compound effect takes time - accounts that post consistently for six months almost always outgrow accounts that post sporadically over two years.

The B2B Angle: Using X as a Lead Channel

If you're reading this as an agency owner, founder, or B2B sales professional, the algorithm question is really a business question: can X be a reliable source of inbound leads and brand authority?

The answer is yes - but only if you treat it as a platform for building specific expertise, not broadcasting. X rewards specificity. An account that posts consistently about one narrow topic for a focused audience builds compounding reach. An account that posts about "everything" builds nothing. Pick two or three topics where you have genuine expertise and own them. The SimClusters system literally cannot find your audience unless it has a clear model of what you're about.

When your content on X starts generating attention - replies, follows, profile clicks - those are people who already trust you before they ever visit your site or see your offer. That's warm lead territory. The closing tool at that point isn't social media tricks, it's good outbound fundamentals. I break down the full system in the Purpose Framework if you want to connect your content strategy to actual revenue.

For those running outbound alongside their content strategy - if you're using X to identify potential clients (by who engages with you, who's in your niche, who's following your competitors) and then taking that outreach off-platform to email or DM, you'll want a solid way to find their contact information. This email finding tool bridges the gap between social discovery and actual outreach - put in a name and company, get a verified email address. And if you're identifying YouTubers or creators through X and want to reach them outside the platform, ScraperCity's YouTuber Email Finder can pull contact info directly for influencer or partnership outreach.

If your B2B prospecting goes beyond just email - cold calling, direct dials, multi-channel sequences - a mobile number finder lets you layer phone outreach on top of your email sequences for the prospects you really want to reach.

For scheduling, drafting, and growing your X presence systematically, Taplio is purpose-built for LinkedIn and Twitter personal brand growth and worth looking at alongside whatever content calendar you're already running.

Common Mistakes That Kill Your Reach

Let me give you the specific mistakes I see most often from B2B operators trying to use X for growth:

Treating X like a broadcast channel. Posting your blog articles and product updates without any conversation-generating content. This trains the algorithm that your account only generates passive likes - and it adjusts your distribution accordingly. Mix broadcast content with conversation-starting content at roughly a 1:3 ratio.

Using the same hashtags on every post. Hashtags on X carry far less weight than they used to - the semantic matching through Grok and SimClusters has largely replaced keyword-based discovery. Two or three relevant hashtags per post are fine; loading up on hashtags looks spammy and may suppress reach.

Posting external links in the body. Still the most common mistake. Every time you post a link directly in your main post on a free account, you're accepting a significant reach penalty. Put the link in the reply.

Ghost-posting. Putting out content and never engaging with the replies. Not only does this hurt your TweepCred, it kills the conversation depth signal that the algorithm specifically rewards. Respond to replies, especially in the first 30 minutes.

Following far more accounts than follow you. The follow/follower ratio is a direct TweepCred input. If you're at 2,000 following / 400 followers, the algorithm is treating you like a low-credibility account regardless of your content quality. Clean up your follow list.

Ignoring your analytics. X Analytics shows you exactly which posts drove the most impressions and engagement. Most people never look at it. Check it weekly, identify which content types drove replies and bookmarks, and build more of that.

What Actually Moves the Needle

To recap - changing the Twitter algorithm in your favor, whether on the consumption side or the growth side, comes down to a handful of non-negotiable moves:

• Feed control: Use "Not Interested" signals, mute aggressively, engage intentionally with content you want to see more of, and use Grok's direct feed customization to accelerate the process.
• Post timing: The first 30 minutes after posting determine whether your content lives or dies. For B2B, Tuesday through Thursday mornings are peak windows. Post when your specific audience is actually online.
• Optimize for conversations, not likes: Reply to your own comments, create posts that demand a response, and remember that a two-way conversation thread is worth approximately 150x a passive like.
• Bookmarks signal value: Create reference-worthy content - frameworks, data, step-by-step guides - that people save. Explicitly invite bookmarks.
• Ditch external links in the main post: Put links in replies to preserve reach. This applies especially to free accounts.
• Use native media: Video under 60 seconds, images, and polls all get algorithmic preference over text-only posts.
• Topical consistency: The SimClusters system finds your audience based on what you consistently post about. Random off-topic content breaks that model and suppresses reach.
• Stay consistent: Accounts that post daily for six months almost always outgrow accounts that post sporadically over two years. Consistency is the compounding variable.
• Protect your TweepCred: Avoid follow-for-follow schemes, bought followers, and behaviors that generate blocks and mutes. One bad signal undoes many positive ones.

None of this is magic. It's understanding what the system is optimizing for - conversation quality and time-on-platform - and building your content strategy around those actual goals instead of vanity metrics. Do that consistently, and the algorithm stops feeling like something working against you.

If you want help implementing a full content and outbound strategy that turns social presence into actual sales meetings, that's exactly what I work on inside Galadon Gold.