X (Twitter) Shadowban and Visibility Filtering in 2026: How to Check and Fix It

Your posts still publish. Your follower count hasn’t moved. But the replies you write under big accounts get nothing, your handle stopped appearing in the search box when people type it, and your impressions fell off a cliff overnight without a single post being deleted.

That combination is the signature of visibility filtering, and X is unusual among the big platforms in admitting the mechanism exists. Instagram and TikTok describe theirs in euphemisms about eligibility for recommendation. X named it in its enforcement policy, built an interface label for it, and gave some accounts a status screen that says in plain language that their reach is limited.

What X has never published is the thresholds, the durations, or the appeal timeline. So the work is diagnostic: figure out which state you’re in, because the four common ones have different causes and very different recovery paths. Treating them as one vague shadowban is why people spend three weeks fixing the wrong thing.

What X actually filters, and where the vocabulary came from

Nearly every term people use for this comes from outside the company: search suggestion ban, search ban, ghost ban, reply deboosting. They entered wide circulation in July 2018, after a Vice News report claimed that several prominent Republican accounts had stopped appearing in Twitter’s search auto-suggestions. Twitter answered with a blog post titled “Setting the record straight on shadow banning.” It denied hiding accounts on the basis of political viewpoint, and in the same post confirmed that it ranks content using behavioural signals and ranks some accounts lower in search and in conversations.

That was the tell. The company denied the loaded word and confirmed the practice.

The framing X uses now is more explicit. Under its approach to enforcement, a post that breaks certain rules is not always removed. Its distribution is restricted while it stays live. Some of that content carries a visible notice saying its visibility is limited, with a link to the policy and an appeal path. Not every filtered post gets a label, which is the whole problem. Labelled cases you can act on. Unlabelled ranking penalties are what make people feel they are shouting into a void.

Two things matter before you start testing. First, filtering operates at multiple levels. A single post can be downranked without the account being touched. An account can be downranked in search while its posts still reach existing followers normally. A reply can be collapsed while the same account’s original posts perform fine. These are separate systems, and diagnosing at the wrong level wastes days.

Second, filtering is often temporary and automated. A large share of what people experience as a permanent shadowban is a short-lived spam-classifier response to a burst of behaviour: a follow spree, a batch of scheduled posts firing at once, a cluster of replies carrying the same link. Those decay. Policy-driven restrictions attached to a violation do not decay on their own; they need an appeal or a serve-out.

Four filtering states, plus one you can actually see

Read the symptom column first and find the row that matches what you’re seeing.

One caveat on that last column: X has never published a duration for any of these states. Those ranges are what account owners report.

The last row is the one people overlook and the one to check first: eleven seconds, and it either ends your investigation or narrows it sharply.

Check the account status screen before anything else

X exposes an account status page in the app and on web: open Settings and privacy, then Your account, then Account status. If your account is carrying an enforcement action (a reach restriction, or a policy violation X has recorded against you), this is where it surfaces, usually naming the policy involved and offering an appeal button.

A clean screen here does not prove you’re unfiltered. Automated spam-classifier downranking generally produces no entry. But a non-clean screen saves you a month of guessing, and that appeal button is the strongest move available to you, because it routes into review rather than into a generic support form that goes nowhere.

While you’re in settings, check two adjacent things. Under Privacy and safety → Your posts, see whether the option to mark your own media as potentially sensitive is switched on. If it is, every image and video you publish sits behind a warning interstitial and drops out of a chunk of recommendation surfaces. People switched it on years ago and forgot. The account-level version of the same flag is worse: ordinary posts with a stock photo get hidden behind a content warning. It suppresses reach hard, quietly, and is appealable.

How to test for a search suggestion ban

The mildest state, and the easiest to confuse with nothing at all.

Get a second account that does not follow you: a colleague’s, or a personal one you don’t use for the brand. Following changes how suggestions rank, so the tester has to be a stranger to you.

Have them tap the search field and type your handle one character at a time. An account with any meaningful following usually surfaces in the dropdown well before the handle is finished. If they type it complete, character-perfect, and you still don’t appear (though you do appear once they press enter and open the People tab), that’s a suggestion ban.

Repeat with your display name rather than your handle. The two are indexed differently and one can be suppressed while the other isn’t.

Suggestion bans are the least damaging state. They’re often triggered by a rapid profile-field change or a new account posting at volume too early, and they lift on their own more reliably than anything else here.

How to test for a search ban

Search bans matter more: they remove you from search results themselves.

The old test, logging out and searching your handle, is now unreliable, and you should know why before you trust it. X restricted what logged-out visitors can see, and search in particular sits behind a login for most sessions. An empty logged-out result tells you almost nothing.

Do it this way instead.

1. Pick a recent post containing an unusual phrase, something specific enough that it appears almost nowhere else. Six or seven distinctive words.
2. From a logged-in second account that does not follow you, paste that phrase into search and press enter.
3. Switch to the Latest tab. The Top tab is ranked and will hide plenty of legitimate posts for reasons unrelated to filtering.
4. If nothing appears, run the query again with the operator from:yourhandle plus one word from the post, forcing X to look at your account’s index specifically.
5. Finally, search from:yourhandle on its own and check Latest. A healthy account returns a reverse-chronological wall of its own posts.

If step 5 returns nothing, or only posts older than a certain date, you have a search ban and you know roughly when it started. That date is the most useful thing you can get: go look at what you did that week.

Run the test twice, several hours apart. Indexing is not instant; a post from ninety seconds ago missing from Latest is normal.

How to test for a ghost ban and reply deboosting

Test these two together. One procedure catches both, and the difference between them is a matter of degree.

Reply to a public post from a reasonably active account you don’t already interact with, so no prior engagement history inflates your position. Wait ten minutes.

Now open that same thread from your second account and scroll the replies. Three possible outcomes:

• Your reply appears in the normal flow. No ghost ban, no meaningful deboost on that thread.
• Your reply appears only after you click “Show more replies” or expand a section warning about offensive content. That is reply deboosting. Your reply exists and is reachable, but it sits below a fold that most readers never open.
• Your reply does not appear at all, at any depth, but is visible on your own profile and in your own view of the thread. That is the ghost ban state, and the one that makes people believe everyone is ignoring them.

The self-view discrepancy is the diagnostic core. X shows you your own content regardless of how it is being distributed, so every check from your own account looks fine. That is not reassurance. It is the mechanism working as designed.

The impressions comparison that shows severity

Open the analytics on individual posts using the small bar-chart icon beneath each one. The count shown under a post is labelled views; the panel behind that icon breaks it out as impressions, engagements and profile visits. Pull the impression count for your last twenty original posts and your last twenty replies, kept in date order.

You want the ratio between the two, and how it changed.

On an unfiltered account, replies to bigger accounts pull impressions in the same order of magnitude as your originals, and often far more. That is the whole mechanic behind reply-based growth: you are borrowing an audience.

On a reply-deboosted account, the pattern inverts and stays inverted. Originals hold roughly steady, because followers still get them delivered. Replies collapse to a couple of dozen impressions each regardless of whose post you replied under. No reply is written badly enough to earn twenty views beneath a post that got hundreds of thousands. This is a distribution problem.

Note the date the ratio broke, then look at your activity in the seventy-two hours before it. Usually the cause is sitting right there.

Two calibration notes. Reply impressions are naturally low when you arrive late to a cooled thread, so only compare replies posted within the first hour of their parent. And one bad week is noise; wait for a break that holds ten to fourteen days.

What actually triggers filtering

Causes divide into behavioural signals, which are automated and usually temporary, and policy signals, which are enforcement actions and usually need an appeal.

Aggressive automation and third-party posting tools. The most common cause among people running a serious content operation, and the most frustrating, because the tools are legitimate. The pattern is the problem. Twelve posts landing at exactly the top of the hour, every hour, from an API client with no manual sessions in between, is indistinguishable from a bot.

Rapid follow and unfollow. Following 200 accounts in an afternoon and unfollowing 180 of them two days later is one of the oldest patterns X’s spam systems detect. The threshold isn’t published, but churn is what gets flagged, more than raw volume.

Repeated reports from other users. A cluster of reports in a short window, including ones later found baseless, can produce a precautionary reach reduction while the account is assessed. If something you posted drew a brigade, that is likely your cause, and waiting genuinely works here.

Sensitive-media classification. Either you enabled the setting, or a classifier applied it to specific posts, or the account carries the flag. Any of the three cuts reach on media posts hard.

Link-heavy patterns. An account whose posts are overwhelmingly outbound links (same domain, shortened, dropped into replies to strangers) matches the spam profile precisely. Put the substance in the post and the link in a reply. Safer, and it reads better.

Duplicate and near-duplicate posting. Publishing identical text repeatedly, or the same text across a cluster of accounts you control, falls under the platform manipulation rules. That includes the popular tactic of reposting your best performer verbatim every few weeks. Rewrite it instead.

Reply patterns that read as engagement farming. Short, generic, high-frequency replies are a recognised spam signature: the same handful of phrases sprayed across dozens of large accounts. This is where a lot of growth advice quietly gets people filtered. If you pair organic replying with paid distribution through LitFame or any other provider, keep the organic side varied; the paid side gives you no cover when your own behaviour trips the classifier.

A recovery plan that matches the state you’re in

Separate the two categories first. If Account status shows an enforcement action, appeal it; everything else is secondary. If the screen is clean and your testing points to behavioural filtering, you are waiting out a classifier, and your job is to stop feeding it.

Day 1: stop the bleeding. Go to Settings → Security and account access → Apps and sessions → Connected apps and revoke anything you don’t recognise or haven’t used in three months. Old growth tools you authorised in 2021 are still acting on your account’s reputation. Pause all schedulers. Stop following and unfollowing entirely.

Days 1–3: fix the obvious content problems. Delete posts carrying visibility-limited labels, and appeal any you think were wrongly labelled; editing will not clear a label, and the edit window is short and Premium-only anyway. Remove duplicates. Take down posts that are nothing but a shortened link. Don’t mass-delete your archive. There’s no evidence it helps, and a deletion spree is itself an anomalous pattern.

Days 3–14: post like a human being who likes the app. Two to four original posts a day, published manually, at irregular times. Real replies written for the specific post. No links for the first week if link patterns featured in your diagnosis. Spend time in the app: scrolling, opening threads. An account that only appears via API calls looks different from one that gets used.

Days 14–30: re-test and read the trend. Run the same search and reply tests from the same second account and compare. Recovery usually shows up first as replies climbing back above the fold, then search results returning, then impressions normalising. It is rarely one moment where everything switches back on.

Now the honest part, which is inconvenient for anyone selling growth services, this site included.

X publishes no fixed duration for any of this. There is no documented 30-day timer and no support channel that will tell you when a behavioural downrank expires. Anyone quoting an exact number is guessing. What can be said honestly is that these are classifier states that relax as the behaviour feeding them stops, so the clock restarts every time you resume the pattern. Going by what account owners report, mild suggestion bans tend to clear inside a couple of weeks of ordinary use, while search bans and reply deboosting run several weeks longer. A minority of accounts carry a low-grade reply penalty for months with no fixable cause anyone has found. That group is real.

If you’ve run a clean thirty days with no movement on any test, the realistic options narrow to appealing again through Account status, or building a second account properly and treating the first as a legacy asset. Nobody wants that answer, but it beats six months of posting into a filter.

Things that do not fix it

Deactivating and reactivating does not clear a filter. It does risk data loss and, with bad timing, permanent deletion.

Changing your handle does not reset your reputation; that attaches to the account itself rather than the name on it. All a rename breaks is your search indexing and your inbound links.

Third-party shadowban checkers are largely unreliable now. Most were built against API access that has since been restricted or priced out of reach. A tool reporting a search ban when a manual from:yourhandle query returns your posts is simply wrong.

Buying a block of engagement to push through a filter does not work either, and it is worth being blunt about why. Distribution services can add visible social proof, which has real value on a healthy account. But volume is not the constraint when the ranking layer is holding your content back. If you’re weighing paid distribution while filtered, fix the filter first, confirm with the reply test that your content reaches the normal flow again, and only then create an account with a provider and layer distribution onto content that can actually travel.

Staying out of it once you’re back

Vary your posting times rather than locking to a schedule. If you use a scheduler, jitter the times by ten or fifteen minutes and mix in manual posts. Keep following slow, and unfollow gradually rather than in batches.

Write replies that could only have been written for that specific post. This is the strongest protection you have, because template detection is what catches most reply-strategy accounts, and specificity is the one thing a template cannot fake.

Keep link posts well below half your output, rewrite recurring content instead of reposting it, and check Account status once a month the way you’d check a billing page. A restriction caught in week one is far easier to appeal than one you notice after ninety days of decline. Run the reply-fold test quarterly even when nothing seems wrong. It takes four minutes, and it’s the only early warning system you have. And if reach matters commercially, write down your own impression baseline before you layer anything on top of it (paid distribution from LitFame included), because you cannot tell what a campaign did, or what a filter did, without a record of normal.

Visibility filtering rewards patience over cleverness. Most accounts recover by doing less, more consistently, for a month. The ones that don’t usually changed tactics every four days and never gave a single change time to show a result.

Frequently Asked Questions

Does X officially admit that shadowbanning exists?

X has denied the specific term while confirming the underlying practice. The company has stated it does not hide accounts based on political viewpoint, but it has also publicly described restricting the distribution of rule-breaking posts rather than removing them, and it applies visible labels to some content indicating limited visibility. The mechanism is acknowledged. What has never been published is which behaviours trigger it, at what thresholds, or for how long restrictions last.

Can I check for a shadowban while logged out?

Less reliably than you could a few years ago. X restricted what logged-out visitors can access, and search in particular is largely gated behind a login, so an empty logged-out search result tells you very little. Logged-out browsing still works reasonably well for checking whether a specific reply appears in a thread. For search testing, use a logged-in second account that does not follow you, and always check the Latest tab rather than Top.

How long does a shadowban on X last?

X publishes no fixed duration, and no support channel will give you a date. These are automated states that ease as the behaviour driving them stops, which is why the clock restarts when you resume the pattern. Going by what account owners report, mild search suggestion bans often clear within about two weeks of ordinary use, while search bans and reply deboosting commonly run several weeks longer. A minority of accounts carry a low-level reply penalty for months. Treat any source quoting an exact number of days as guesswork.

Why do my replies get fewer impressions than my original posts?

On a healthy account that ratio usually runs the other way, because replies borrow the audience of the post you replied under. When replies consistently pull a fraction of your originals, and the gap holds for two weeks or more regardless of whose post you reply to, that points to reply deboosting. Confirm it by opening the thread from a second account and checking whether your reply sits below a show-more-replies fold.

Will deleting old posts remove a visibility restriction?

There is no evidence that mass deletion lifts a filter, and a sudden deletion spree is itself an unusual behaviour pattern. Deleting specific posts is worth doing in narrow cases: posts carrying an explicit visibility-limited label, exact duplicates you published repeatedly, and bare link posts if link patterns were part of your diagnosis. Beyond that, leave your archive alone and focus on changing current behaviour instead.

Do third-party shadowban checker tools still work?

Most are unreliable now. They were built against Twitter API access that has since been restricted or priced far higher, so many return cached, inferred, or simply incorrect results. Manual testing is more trustworthy: search an exact phrase from a recent post using a second account, run a from:yourhandle query on the Latest tab, and check whether your replies appear in threads for someone who does not follow you.

Can buying engagement cause a shadowban?

The pattern matters more than the purchase. Sudden implausible spikes, engagement from obviously inactive accounts, and identical activity landing across many posts at once all resemble the platform manipulation signals X’s systems look for. Gradual delivery on content that already gets some organic response is far less conspicuous. If your account is currently filtered, resolve that first, because added engagement does not override a ranking penalty applied further up the pipeline.

Does changing my handle reset my account’s reputation?

No. Reputation attaches to the underlying account identifier, not the display handle, so renaming carries any existing restriction across with it. What a handle change does accomplish is breaking your search indexing temporarily, invalidating inbound links and mentions, and confusing people who knew you by the old name. If you have already exhausted appeals and a full clean month of normal behaviour, a genuinely new account is the option to weigh.