Elon has revealed how the X algorithm works i've studied it & figured out how to make it work for me THESE PAST 9 DAYS I'VE: > HAD 7 MILLION IMPRESSIONS > GAINED 7k FOLLOWERS in this thread, you'll learn how to: > go viral > avoid the shadowban > make the X algo work for you
in a nutshell: > reply to replies - amplifies posts > engage a lot > get people to visit your profile > DON'T get blocked or muted > tweet & reply often > tweet relevance drops 50% every 6h > communities posts makes it easier to link you to group audiance > blue check mark
this will be a long but INVALUABLE read, you wanna go through it in full (if you're serious about X) some key topics: > influence of follower type & engagement signals > importance of replies (especially to replies on your own tweets) > the impact of blocks/mutes > the algo
X's algorithm is engineered to reward specific actions some positively > replies, retweets, and likes and some negatively > mutes & blocks basically, you either go VIRAL or get SHADOWBANNED
your visibility is decided by several factors > pagerank-style trust > reputation graphs > live engagement scores > content bundles > safety filters the X's algo tracks every click, block, mute, reply, and share so, how does it all work together?
X's simclusters how does X show your tweet to people who don't follow you? simclusters > it groups users into interest-based communities > then shows you posts popular in your clusters > the algo doesn't care who follows you > it cares who engages in a similar fashion to you
to show up on someone's For You page, your tweet has to go through 1. light ranker: a fast pre-filter algorithm, a `maybe` or `no` decision maker 2. heavy ranker: a deeper neural network that scores tweets using thousands of features to predict engagement likelihood
the ranking system works in three stages > sourcing tweets, from people you follow and others you don't > scoring them using massive neural networks > mixing everything into a scrollable feed each of those steps affects how, if at all, your tweets travel
X’s reputation system: the Tweepcred the Tweepcred is a pagerank-like system to evaluate how credible you are > your behavior on X gets you a Tweepcred `reputation` score > it looks at who interacts with you and how often > this impacts how far your tweets travel
if you break the algo rules > your score goes down & reach disappears what are the rules of the X algo? > post `offensive text`? ~50-80% reach reduction > offensive username? ~50-80% reach reduction > all caps tweet? seen as `shouting`, huge penalization > links? penalized
the X algo comes with its own blacklist of `bad words` & topics > trigger them & your reach disappears we don't get get to see them, they're hidden in text files > `adult_tokens.txt` > `offensive_topics.txt` > `offensive_substrings.txt`
there are other-weird-ways to get deboosted > low quality text (misspellings/tWEEtiNG LiKE tHiS) > replying to people who don’t follow you > spamming new accounts start at `-128` Tweepcred and you need a minimum score of 17 to be seen next-up: followings:followers ratio
followings:followers ratio this impacts your reputation score heavily try to maintain a followings:followers ratio ~1:1.67+
if you follow way more people than people follow you back, your reputation score takes a massive hit > the higher your following-to-follower ratio, the worse your reach gets why though?
> stops follow-for-follow spam & fake engagement > following people is now less rewarding for everyone > limits bots & users who follow tons of people just to get attention now, let's talk about how the X algo shadowbans you
how does shadowbanning work? > your account gets a label that destroys your reach some of the labels that can be applied to you > mentioning a person several times > untrusted(LOL) urls > low quality tweets > toxicity > spam > nsfw > gore
if a tweet is getting > mutes > blocks > reports > not interested clicks > see fewer posts like this clicks the algorithm actively downranks it - these are negative signals, and they stick (potentially DEBOOSTED FOR 3 MONTHS) X's algo: too many negative signals = invisibility
alright, let's move on to how we can get BOOSTED the basics: > if people spend 2+ secs looking at your text tweets, it’ll be pushed further > if people spend 10+ secs watching a video tweet, it'll get boosted big time
on your tweets, if you're getting > replies > retweets > likes > bookmarks > follows after someone sees your tweet > profile clicks > shares > video views, especially long ones the algo pushes you way up — these are strong positive signals
> replies to replies: heavily weighted (explains why threads spiral HARD) > if your reply to a reply triggers more replies, you’ve entered `replyguymaxxing` > replyguy chains help define the simcluster > replying to replyguys is a growth vector is the age of reply guy back?
and it's not just you, it's who interacts with you > the algorithm uses `realgraph` to model relationships > if accounts that follow and engage with you also engage with someone else, that person’s tweets may be shown to your followers it's contagious reach
among the ways you get boosted: > `blue check mark`: critical, without it YOU CANNOT make it on X > `legacy verified`: extra nice boost > `viral account`, or being considered one, boosts you also lesson: go viral in order to go viral LOL
paying for the blue check sets your reputation score to 100 > new accounts start at -128, so you're basically toast without it then the algorithm hits you with > following-to-follower ratio adjustment > and also factors in things like your device and how old your account is
X's diversity filters why doesn’t your feed get spammed by 1 person or topic? diversity filters > they cap how many tweets from the an author or subject can show up > features like "author diversity" and "slop score" keep things fresh > variety=more scrolltime=more ads=more X
to avoid repetition > X limits how many tweets from a single author appear back-to-back > so if you're tweeting in bursts, spacing them out might help you hit different segments of the timeline instead of getting filtered out
diversity filters ensure that you don't keep getting the same stuff this is important when a major event happens for example
X's virality if your tweet lands in > video carousels > pinned modules > bookmark modules > trends then congrats > X's mixer boosted it as part of a content bundle > and you just earned premium shelf space
> don’t get muted or blocked > avoid offensive language, it tanks reach > reputation score is everything, protect it > post stuff that takes >2s to read > quality > spam, esp. for media (10s watch time rule) > engagement matters: quotes + bookmarks hit hardest > blue check helps
tl;dr > if your tweets spark real engagement, earn trust, and get people talking-not just liking-you'll ride the algorithmic wave > X has baked in reputation, content understanding, safety filters, and real-time feedback to decide what blows up tweet accordingly
where to find the algo? https://github.com/twitter/the... p.s. use the analytics dashboard
FAQ #1 engagement on an old post? > a single reply acts as a signal boost and reactivates it











