X has spent 2026 making its reach penalties legible. It open-sourced the ranking code, shipped an "Under the Hood" page that shows the visibility labels on your own account, and in September expanded that page to reveal when a post was withheld in a country under local law. Put together, the picture is clearer than it has ever been: reach on X is not a single number the algorithm hands you, it is a base score the recommendation system computes and then reduces every time your post trips one of a specific set of penalties. This guide catalogs all of them — the content labels that demote a post on merit (an "offensive" classification, spam, toxicity, NSFW, untrusted URLs), the off-platform link demotion that creators have argued about for years, the negative-reaction signals (mutes, blocks, reports, "not interested") that are applied per viewer rather than platform-wide, the Community Notes and engagement-bait penalties, and the legal withholds that remove your posts in an entire market on grounds you cannot write your way out of. It explains how the penalties are actually applied — the difference between an account-level label that shadows everything you post and a per-user signal that only costs you that one viewer — how to check which ones are hitting you now that X shows you, and, most usefully, how to stop minting posts that trigger them in the first place. Because the honest conclusion is that avoiding penalties is necessary but not sufficient: a feed that can cap your reach on quality labels you partly see and legal demands you see only after the fact is a feed you should never let be the single gate on your audience.
The most useful mental model for X reach in 2026 is subtractive. The recommendation system — the Grok-powered ranker X now calls Phoenix, which replaced its older hand-tuned heuristics earlier in the year — computes a candidate set of posts for each user's For You feed and scores every post on how likely that user is to take more than a dozen actions on it. That score is your reach potential. Then, before the post ever competes for a slot, a set of penalties can reduce it: a content label, an off-platform link, a pile of negative reactions, a Community Note, or a legal withhold. Understanding X reach is mostly understanding that stack of penalties, because the boosts (replies, shares, dwell, video completion) are well documented and comparatively intuitive, while the penalties are where reach quietly disappears without a notice.
This is not guesswork anymore. X open-sourced its recommendation code, and independent analyses have put numbers on both the boosts and the penalties. More importantly for a creator, X spent 2026 making the penalties visible to the person they hit: the August "Under the Hood" page showed users the reach-limiting labels on their own account for the first time, and the September 19, 2026 expansion added a category for posts withheld in a country under local law. The primary-source write-up of that transparency expansion has the details; this guide is the full catalog of what those labels and signals actually are, how each one is applied, and how to stop minting posts that trigger them.
The penalties fall into five groups, and they differ in a way that matters enormously for how you respond: some are avoidable by writing better, some are per-viewer and only partly in your control, and one is entirely outside it. Confusing them is how creators end up either paranoid about things they can't change or complacent about things they can.
The most familiar penalty is a content classification that reduces a post's distribution. The open-sourced code documented an "offensive" text label historically cited as a roughly 80% reach reduction, and the broader set of labels that suppress distribution includes spam, toxicity, NSFW, gore, and untrusted URLs. The critical distinction inside this group is scope. A single flagged post is demoted on its own. But some labels attach to the account — repeated toxicity, spam behavior, or persistently untrusted links can shadow everything you post, not just the offending tweet, which is the closest thing to the "shadowban" X refuses to call it. This is the category the Under the Hood transparency tool was built to surface, because an account-level label is otherwise invisible: your posts just underperform with no explanation.
For years the standard advice was that adding an external URL to the body of a post cut its reach, and the near-universal workaround became posting the content natively and dropping the link in a reply. In July 2026 Elon Musk said X had not penalized link posts "for over a year" — but independent testing through the year kept finding that link-in-body posts received dramatically fewer views than equivalent link-free ones. The honest status is unsettled: leadership says there is no penalty, measurements say link posts underperform, and the gap may be an explicit flag or simply the ranker favoring engagement that happens on-platform. Either way the practical outcome is identical, and the full breakdown of the link-penalty debate lands on the same default this guide does — keep the value in the post and put URLs in a reply until the data agrees with the statement.
This is the group most often misunderstood. Analyses of X's current code attach heavy negative weights to hostile signals — a report is worth a large negative score, a block and a mute substantial ones, and a "not interested" tap a smaller but real penalty. The number that matters, though, is not the size of the weight but the scope: these penalties are applied per user, not platform-wide. One person muting or reporting you damages your reach with that person, not across the whole platform. That reframes the advice completely. You are not trying to achieve zero negative reactions — that is impossible and chasing it makes you bland. You are trying to avoid the content that generates negatives at a rate high enough to matter across your audience, which is a very different and more forgiving target. The full read of X's per-engagement point weights shows why shares and replies, not the avoidance of every mute, are the real lever.
Two narrower penalties round out the on-merit group. A post that receives an active Community Note can see its distribution reduced while the note stands, which is the system working as intended — a correction should not travel as far as the claim it corrects. And engagement-bait — "like if you agree," "reply YES for the link," reply-gating, and manufactured-interaction phrasing — is demoted, because X's detection for artificially farmed engagement has tightened. The lesson is that gaming the boost signals directly is itself a penalty trigger: the ranker rewards genuine replies and shares and demotes the attempt to manufacture them.
The newest and most different category is the one the September 2026 Under the Hood expansion surfaced: a post withheld in a specific country following a legal demand under local law. This is not about the quality of what you posted. It is a jurisdictional removal — your content is hidden from a whole country's feed on grounds that have nothing to do with your craft and that you typically learn about only after the reach is gone. There is no writing-better fix, and the transparency page names the country but does not reverse the withhold. It is the clearest proof that reach on X is conditional on things entirely outside your control, and it is the reason the last section of this guide argues that avoiding penalties, while worth doing, can never be your whole strategy.
The single most important structural fact is the split between penalties that follow the account and penalties that follow the viewer. An account-level label — earned through repeated toxicity, spam, or untrusted links — shadows everything you post until it clears, so it is worth taking seriously and worth checking for in Under the Hood if you are eligible. A per-viewer signal, like a mute or a block, only costs you that relationship and cannot on its own tank a healthy account. Creators routinely invert these: they panic over a handful of mutes (per-viewer, minor) while ignoring a pattern of borderline posts that risks an account-level label (systemic, severe). Diagnosing which kind of penalty you are actually hitting is the first practical step, and it is exactly what the companion how-to on checking for and avoiding X reach penalties walks through in order.
Layered on top of both is the boost side of the ledger, which matters because reach is the base score minus penalties — so a post with strong positive signals can still out-distribute a lightly penalized one. Replies are weighted far above likes, an author-answered reply higher still, and shares — passing a post along, on or off the platform — sit at the top of the positive signals in the current code. Early engagement velocity compounds all of it: the first 30 to 60 minutes disproportionately determine a post's eventual reach, so the same post published into a dead window underperforms one published when your audience is active. None of this is a penalty, but all of it is why penalty-avoidance alone is not a reach strategy — you also have to earn the boosts.
The avoidable penalties reduce to a short discipline. Keep posts native — the value lives in the post itself, and any URL goes in a reply, which sidesteps the link question regardless of how it resolves. Write for conversation rather than passive taps, because replies and shares are the signals that lift reach, and posts built to be argued with or forwarded out-distribute posts built to collect likes. Stay clear of the content that earns account-level labels: no manufactured engagement-bait, no mass-mentioning of unrelated handles, no all-caps shouting, no borderline-toxic framing that invites reports at scale. And post into your audience's active window so early velocity works for you instead of against you. The tactical guide to posting on X for maximum reach works these into a repeatable routine, and the brand-presence playbook for X sets the account-level habits that keep you out of the shadow-label territory in the first place.
What this discipline cannot do is touch the two categories outside your control — the per-viewer negatives that some people will always generate, and the legal withholds that a jurisdiction can impose regardless of what you wrote. That is not a reason to ignore the avoidable penalties; a clean, native, conversation-driving account is genuinely more resilient. It is a reason to be clear-eyed that even a perfectly-behaved account is still subject to a feed that can cap its reach on grounds it partly sees and grounds it sees only after the fact. Which leads to the part of the strategy that penalty-avoidance can never cover.
The honest conclusion of the catalog is that reach on X is only partly yours to control, so the strategy has two halves: stop minting posts that trip the avoidable penalties, and stop depending on a single feed for the reach that penalties — avoidable or not — can take away. Kompozy is built for both halves, and its angle on the first is the one worth pressing here: it governs the penalty triggers out at the moment of generation, before a post exists, rather than catching them in review after you have already written something risky. The Persona Brief encodes your voice, your phrasing, and — critically for X — a banned-word and prohibited-topic filter, so every post it generates is produced clean of the language that earns an "offensive," toxicity, or spam label. You are not proofreading penalties out one tweet at a time; you have configured them out of the whole output stream.
Because Kompozy is a generation-and-publishing engine, it produces platform-native assets by default — captioned Persona Shorts, brand-exact Carousel Posts, Clipped Shorts, quote cards, and text posts across 18 formats — none of which need an external URL in the body to deliver their value, which sidesteps the link penalty by construction rather than by remembering to move the link to a reply. And it publishes on Autopilot behind a per-post review gate, so you schedule into your audience's active window for the early-velocity boost and still approve every post before it ships. The banned-word filter and native-format generation between them cover the on-merit penalties this guide catalogs, at the source, across every post rather than one at a time.
The second half is the one no amount of clean writing can solve on X alone: legal withholds and any opaque account-level label cap your reach on that feed no matter what. The structural answer is multi-platform distribution — from one source, Kompozy fans the same idea across the eight social platforms plus blog and email, so a post withheld in a country on X, or an account quietly demoted there, is one you are already reaching through Instagram, TikTok, YouTube, LinkedIn, Threads, your own blog, and your newsletter. State the limit plainly: Kompozy does not give you an X reach report — Under the Hood is X's own tool for that, and you should use it if you are eligible. What Kompozy removes is the part you can actually engineer away — the penalty triggers baked into how a post is written, and the single-platform dependence that turns any one feed's penalty into a reach problem you have no hedge against.
Reach on X is a base score the recommendation system computes for each post, then reduces when the post trips a penalty. The documented penalties fall into four groups. Content labels demote a post on merit — an "offensive" text classification (historically cited as roughly an 80% reach cut in the open-sourced code), plus spam, toxicity, NSFW, gore, and untrusted-URL labels, some of which shadow an account rather than a single post. Off-platform links have long been demoted, though X's leadership now disputes an active penalty while independent tests still show link-in-body posts underperforming. Negative reactions — mutes, blocks, reports, and "not interested" taps — carry large negative weights but are applied per viewer, so one person muting you costs you that person, not the whole platform. And legal withholds remove a post in a specific country under local law. Treat the exact multipliers as directional: they come from open-sourced code that X re-tunes without publishing every current weight.
It is the enforcement philosophy X published in 2023. Instead of removing a rule-bending post outright, X may reduce its distribution — excluding it from search, trends, recommendations, and the For You and Following timelines while leaving it visible on the author's own profile. It is the framework under which most reach penalties operate: your post is not deleted, it is quietly demoted. The 2026 open-sourced ranking code and the Under the Hood transparency page make which posts carry those reach-limiting labels far more visible to the author than they used to be.
Since August 2026, X's Under the Hood page shows eligible users the visibility labels applied to their account and posts over the prior month, and as of September 19, 2026 it also shows whether a post was withheld in a country under local law — naming the country. Eligibility is gated to accounts at least a year old that posted actively (around ten or more posts) in the prior month, so not everyone can see it yet. If you can't, the older signals still apply: a sudden, sustained drop in impressions on posts that previously performed, your replies not appearing under a target account's tweet, and your posts missing from search for your own handle are the classic symptoms.
The status is genuinely unsettled. In July 2026 Elon Musk said X had not penalized link posts "for over a year," but independent testing through 2026 has repeatedly shown posts with a URL in the body receiving dramatically fewer views than equivalent link-free posts. Whether that gap is an explicit penalty flag or an indirect effect of the algorithm favoring on-platform engagement, the practical outcome is the same: link-in-body posts tend to underperform. Until the measurements agree with the statement, the low-risk default remains keeping the value in the post itself and putting any URL in a reply.
You can avoid the ones you control and not the ones you don't. Content-label, link, and engagement-bait penalties are avoidable by making clean, native, conversation-driving posts and keeping links out of the body. Per-viewer negative signals you influence but cannot eliminate — some people will always mute or scroll past. And legal withholds are entirely outside your control: a country-level demand removes your post regardless of how good it is. That last category is why avoiding penalties is necessary but not sufficient. A feed that can cap your reach on labels you partly see and legal demands you see only after the fact should never be the single channel your audience depends on.
Reach on X is a base score the recommendation system computes for each post, then cuts whenever the post trips a penalty: content labels (an "offensive" classification, spam, toxicity, NSFW, untrusted URLs), off-platform link demotion, per-viewer negative signals (mutes, blocks, reports), Community Notes, engagement-bait, and legal withholds that remove a post in a specific country. Some labels shadow the whole account; negatives are applied per viewer. Since 2026, X's Under the Hood page shows eligible users which labels and withholds hit them — but seeing a penalty explains lost reach, it does not restore it.
Get started → · ← All guides · Compare Kompozy vs other tools