Am I shadowbanned on X? How to check, from X's own code
A Twitter shadowban, now on X, is a label that hides your posts from people who don't follow you. We read X's published code for every rule that does it: where it hides you, how long it lasts, how to check, and which labels X's own Under the Hood report leaves out.
TL;DR
- Labels can hide you in For You, in reply and mention notifications, and in search. Search suggestions don't appear in the published code.
- Under the Hood shows the For You labels but not those for search and notifications. It is a pilot and needs an account at least a year old with 10 or more posts in the past month.
- Accounts with a reputation score of 54 or more, or over 25,000 followers and no score, skip most spam rules, but the engagement-bait label applies to everyone.
- Asking for likes, reposts or follows earns a spam label with no end date, while many automated labels end after 7 or 30 days.
- We found no published rule that labels a post because a scheduler sent it, but duplicate content blocks reply notifications, so vary the text you schedule.
On this page · 9 sections
- Where a shadowban hides you on X
- For You: the labels Under the Hood shows
- Replies and mentions: labels Under the Hood leaves out
- Search: two labels, no report
- How to check if you’re shadowbanned on X
- How long does a shadowban last on X?
- How to get unshadowbanned on X
- Does scheduling get you shadowbanned?
- When it isn’t a shadowban
X’s code never uses the word “shadowban”, but it does what people mean by it. A label on a post or an account can hide it from people who don’t follow you, while you and your followers see nothing wrong. We read the November 7, 2026 snapshot of X’s published code and traced every rule that does this: where it hides you, what triggers it, how long it lasts, and whether X tells you.
X’s Under the Hood report lists the labels that hide you in For You. It leaves out the labels we found for search and for reply and mention notifications.
Where a shadowban hides you on X
X ranks posts and decides visibility in separate systems. A post can score well and still be dropped by a visibility rule because of a label. Most of these rules apply only to people who don’t follow the author, and X’s README says so directly: “The same post is allowed to a follower.”
| Where | What can hide you | In Under the Hood |
|---|---|---|
| For You | Spam, malicious-link, adult-content, abuse, impersonation and compromised-account labels | Yes |
| Reply and mention notifications | Most For You labels, more account and post labels, and the other person’s own filters | Only the labels shared with For You |
| Search | A search blacklist label and a spam label on posts | No |
| Search suggestions | Nothing in the published code | No |
For You: the labels Under the Hood shows
For people who don’t follow you, X drops posts from accounts labelled SpamHighRecall, AbusiveHighRecall, DoNotAmplify, compromised, read-only or impersonating, and posts labelled SPAM_HIGH_RECALL, MALICIOUS_URL or DO_NOT_AMPLIFY. Every one of them appears in Under the Hood’s label list, which gives the main spam label’s effect as “Post hidden from recommendations to non-followers.”
The published rules show what sets several of them:
| What happens | Label | How long |
|---|---|---|
| A post asks for likes, reposts or follows | Post SpamHighRecall |
No end date |
| X flags the account for AI slop | Account SpamHighRecall |
7 days |
| X’s spam models flag the account | Account SpamHighRecall, sometimes with a check that you’re a live person |
7 or 30 days |
| A post is gibberish or a fast spam reply | Post SpamHighRecall |
30 days |
| You pin a post whose link has a bad or low-quality verdict, or follow someone while it’s pinned | Account SpamHighRecall |
7 days |
| A post links to a site with a low-quality verdict | Post SpamHighRecall |
No end date |
The sources are X’s engagement-bait classifier, its enforcement rules for accounts and posts, and the pinned-link rule.
Accounts with a high reputation skip most of these. In the published rule engine that means a reputation score of 54 or more, or a label predicting a high score. An account with no score and no such label qualifies with over 25,000 followers, and government accounts with a gray check skip these rules too. X’s enforcement service uses thresholds it keeps private. The engagement-bait label applies to everyone.
Replies and mentions: labels Under the Hood leaves out
When you reply to or mention someone who doesn’t follow you, X runs another set of rules before it notifies them. It drops the notification if your account has most of the For You labels or NotGraduated, EngagementSpammer, AgathaSpamTopUser, BlinkWorst or DownrankSpamReply, or if the post is labelled spam, low quality, duplicate content or a risky reply. If they have the quality filter on, an untrusted link, a duplicate mention or a spam-reply label drops it as well. Your reply still exists, but the person never hears about it.
None of those extra labels appear in Under the Hood, and the published rules show what sets only some of them. An account that scores extremely high for spam reports per like, or for a likely spam suspension, gets AGATHA_SPAM_TOP_USER for 7 days. A post X flags as AI slop gets a 30-day label that keeps it out of non-followers’ reply and mention notifications.
The other person can filter you too. X’s notification filters can drop mentions and replies from accounts without a confirmed email or phone, with the default profile photo, created recently, or that don’t follow them. A confirmed email, a confirmed phone and a real profile photo get you past three of those filters.
The order of replies under a post is a separate system. X’s reply ranker takes, among other inputs, whether the reply’s author is verified, whether the original poster liked the reply, and a score from Grox, X’s post classifier. A reply can sit low in a thread with no label at all.
Search: two labels, no report
X’s search ranking isn’t in the repo, but its rule engine is, and two published rules label posts for search. A post with a link that resolves to an unsafe verdict gets SEARCH_BLACKLIST along with UNSAFE_URL, MALICIOUS_URL and DO_NOT_AMPLIFY, so it also leaves For You for non-followers. And when an account scores very high for spam reports per like, or for a likely spam suspension, a rule named for search’s top posts labels each new post AGATHA_SPAM for 7 days.
Under the Hood shows MALICIOUS_URL, but not SEARCH_BLACKLIST or AGATHA_SPAM. Search suggestions, where your name appears as someone types, don’t show up in the published code at all, so the code can’t confirm or rule out a “search suggestion ban”.
How to check if you’re shadowbanned on X
- Open Under the Hood. It covers the previous month and lists the labels on your account and posts, what each one does, and how to appeal. X runs it as a pilot for a small group, and it needs an account at least a year old with 10 or more posts in that month.
- Ask someone who doesn’t follow you to search for the exact words of a recent post. If it’s missing, check the post for a link first. Our X shadowban test builds that search and the others for you.
- Reply to someone who doesn’t follow you and ask whether the notification reached them. If it didn’t, the labels and filters above are the place to look.
- Don’t take an automatic checker’s all-clear as proof. Checkers can’t read your labels, and on November 9, 2026 a popular one reported no search ban for an account of ours whose posts X’s own search hid. X’s API still returned every one of them.
How long does a shadowban last on X?
The published code sets 7 days for the AI-slop account label, the pinned-link label and both Agatha labels, 7 or 30 days for other spam-model flags, and 30 days for gibberish and fast spam replies. Engagement bait, low-quality and unsafe links, and some others have no end date in the code we read, and labels a person applies by hand aren’t in it at all.
How to get unshadowbanned on X
Start with the label in Under the Hood and the appeal it shows. When an appeal succeeds, X holds off re-applying that action for a set period.
Remove the cause where you can. Unpin a post with a flagged link, which stops new labels but doesn’t end one already applied, stop asking for likes and reposts, and delete a post that broke the rules: X keeps a read-only account limited “until the post in violation is removed.” Some labels ask for something specific, such as a password reset for a compromised account or a security challenge for an abusive-account label. Many automated labels end on their own after 7 or 30 days, but the link and engagement-bait labels have no end date in the code, so appeal those.
Does scheduling get you shadowbanned?
We found no published rule that labels a post because a scheduler sent it. X’s rule engine receives the app that created a post, but none of the published rules uses that field, and X withholds some rules. We sell a scheduler, so check that against the code. Repeating yourself is a different matter: duplicate content is one of the labels that blocks reply notifications, so vary the text you schedule.
When it isn’t a shadowban
Most drops in reach come from ranking, with no label involved. For You drops posts older than 48 hours, discounts accounts the viewer doesn’t follow, and shrinks your second and third post in the same feed load. How the X algorithm works covers the ranking side, from the weights to the new-author boost.