Building A Quality X Account: Overview
Building A Quality X Account: Overview
Introduction
- We will summarize how algorithms work on modern social media and X in particular
- We will explore what analysts have learned from the public release of some of X’s algorithmic codebase, focusing on what to do, and not do, to boost your account’s reputation
- Following sections will offer suggestions on how you can put your understanding of the algorithms into practice, develop an engagement strategy to take advantage of algorithms, and use your X account in a way that improves its quality and visibility.
Algorithms
- How does a social media platform decide what to show in a user’s feed? Simply showing the posts of all the people a user follows in chronological order leads doesn’t scale well; as users follow tens of thousands of accounts, it becomes hard for a user to find the information they’re truly interested in.
- It also hampers discovery—the process of finding new profiles that may be relevant to a user’s interests.
- The answer is algorithms: automated processes that work to serve up content that keep a user interested and on the site.
Algorithmic Power on X
- X has two tabs for your timeline: Following, which shows posts from accounts you follow, and For You, which also includes posts from accounts that you’re not following. Both are governed by X’s algorithms.
- Alongside determining the visibility of posts, the algorithms also dictate how high up replies will appear (and whether they will be hidden), and how broadly an active Space will appear at the top of the X app.
- Since Twitter, now X, was founded, its algorithm code has been kept secret, but in 2023 this code was made partially public by new owner Elon Musk. Coding experts have now reviewed the code and drawn conclusions about its workings.
- Since Musk’s takeover, he’s promised “freedom of speech, but not freedom of reach,” and has vowed to tackle spam and bots on X. Understanding the tools available on X, understanding how user activity results in wider or narrower reach, and undertaking reach-boosting behavior is the key to growing your account.
- As part of its algorithms X uses two main tools to determine the reach of content: user reputation and its post recommendation algorithms.
- By implementing a strategy that strengthens your reputation and generates posts that the algorithm is likely to recommend, you can increase the likelihood that your posts will be seen by more users.
Recommending Posts
- On their blog. X Engineering went into the mechanics of how they recommend posts to users. With hundreds of millions of posts created daily, X is able to serve up posts that are highly relevant to you.
- When a user loads the For You tab, X tries to extract the best 1500 posts: half from users you follow and half from users you don’t follow.
- To rank the posts from users you follow, X uses a model that predicts the likelihood of engagement between you and a given user.
- To find relevant tweets from outside your network, X looks at what posts people you follow have engaged with recently.
- X also uses algorithms that map out informal communities on X (like the Web3 community); the more users that engage with a post from a community, the more that post will be shown to other members of that community.
- X also ranks posts with an advanced neural network, trained continuously on interactions with posts across the platform, tagging each post with ten labels that give each post a score.
- After ranking, X applies heuristics like filtering out accounts you’ve muted and blocked, showing conversations in context, and social proof (ensuring that someone you follow has engaged with a post or followed its author) as a quality filter.
Reputation
- To help users sift signal from noise, X also scores user reputation, calculating a user’s influence based on their interactions with other users. The system is called “TweepCred,” with a name left over from X’s previous Twitter branding, and has been in place for many years; removing it requires a major refactor of X’s code.
- At the heart of TweepCred is a customized version of the PageRank algorithm, developed by Google to rank search engine results.
- TweepCred uses PageRank to graph X users and their interactions, and assigns a numerical score to each user, based on the number and quality of their interactions with other users. The more a user interacts with high-quality users, the higher their TweepCred score—and the more interactions with low-quality users, the lower the score.
- Over the long term, a high TweepCred score means that your posts will get shown to your followers as well as other users interested in the topics you talk about—giving you continual opportunities to grow your followers and make meaningful new connections.
TweepCred’s Ranking Factors
- When the X recommendations algorithms went public, observers examined the code and drew conclusions about how X calculates TweepCred. These are educated guesses and can always change as the algorithms update.
- Alex Finn, a coder and professional marketer on X who came to prominence as part of the Web3 community, posts regularly updated content about growing on X, based on his analysis of the publicly available X codebase. His ongoing exploration of the algorithm has shed some light on how TweepCred works.
- Finn’s most recent 2024 full analysis is linked in full here.
TweepCred Dings
According to Finn, the top actions that harm your TweepCred are:
- Following significantly more people than follow you. This is evident in the public code, but anecdotal reports are that unfollowing people too quickly can also lower your score, as it looks like bot behavior.
- Interacting (likes, shares, replies) with low-reputation accounts. If you look at a thread on X, some replies are hidden behind a “Show probable spam” button. Don’t interact with these hidden comments or you’ll harm your TweepCred score.
- Blacklisted topics: We don’t know what they are precisely, but the code shows that there’s a penalty for discussing what it calls blacklisted topics. In the same lines of the code are references to offensive content and offensive users, so it’s advisable to use common sense and avoid hateful, controversial and conspiracy topics.
- Posts that loads of people tend to scroll by may be penalized, so avoid short, boring, no-images-or-video posts.
- Hateful speech: Again, these aren’t specified, but the code indicates that offensive language deboosts your posts. Foul or hateful language, attacks, and abuse are likely included.
- Gore, violence, spam (especially crypto spam), sexual content, and short, low-quality posts and replies also lower your TweepCred.
- It your account falls foul of X moderation tools and suffers an administrative penalty, like suspension, search bans and suggestion bans, this will affect your score even after you’re reinstated.
- Anecdotal evidence indicates that it’s far easier to suffer penalties for one mistake than it is to recover your score, so be very careful!
TweepCred Boosts
Finn also highlighted actions that can boost your TweepCred score.
- Engagement from high-reputation posters: Just like interacting with someone from “Show probable spam” can harm your score, getting a like, share or reply from a high-reputation account can boost it dramatically. Meaningful, engaging replies to their posts make this more likely to happen.
- Recent changes indicate that there’s a strong boost to video content—creating original videos can raise your score.
- While controversy should be avoided, talking about trending topics is helpful as the code indicates these can appear in For You even without the regular social proofs. Trending topics show up in X search and change multiple times a day and vary based on your own interests and communities on the platform.
- Avoiding links outside X in your first post of a thread boosts the thread; try to link to a post on X instead or share the outside link in the replies.
- Posts people stop to open and posts that take a while to read are boosted, so threads and long posts that don’t give up the full content immediately, as well as posts with images and video, benefit.
- Having a paid X account boosts your score and makes your replies show up higher, so subscribe to X’s Premium service at the highest level you can afford. Basic, Premium and Premium Plus are available and also offer added X functionality.
- Avoid anything that can get your account suspended, restricted in its reach, or banned, as even when reinstated the score penalties are long lasting.
- A desirable or elite-looking account—many more followers than following, having subscriptions turned on, having an affiliate stamp from a reputable company—is more likely to get high-status engagement and this will boost your score.
- There’s not evidence in the public codebase, but reports indicate that speaking on stage in crowded Spaces boosts your score.
- The most important thing by FAR is there’s no shortcuts to creating quality content—high-signal replies, useful threads, and compelling video will make your account more valuable and more visible over time.
Don’t Game The Algorithms
- The X algorithms are continuously updated in an ongoing arms race with spammers and botters. Don’t try to game the system—you could risk your X account being banned and lose the many hours of work you’ve put into it!
- Musk and X’s shareholders are incentivized to reward real engagement, as the more quality attention goes to X, the more ads they can sell and the more money they make.
- /Reach works on the same principle; mission creators want quality engagement, so that’s what we incentivize. By the same token, any unethical behaviour will be punished—including potential loss of $reach rewards or banning from the platform.
- In the next sections, we’ll list what you should never do, and why. Then we will explore ethical, practical ways that busy people can use to produce effective, real engagement and boost account quality with every action they take on X. By engaging ethically, your account will grow sustainably—and you’ll be well on your way towards increasing your account’s value and influence.
Discussion Questions
- Find your ten favorite big accounts on X. Have a look through their posts and replies. What posts of theirs are getting lots of good engagement? What posts are getting less engagement?
- Find a few busy threads from these big accounts. Can you tell what replies are definitely from real humans, and which might be from bots or spammers? What makes you recognize one or the other?
- Take a look at your own timeline and find your own posts that have gotten more likes, reposts and impressions (views). Make a guess as to why these posts may have been successful. What can you change to get more chances of success?