A real audit of a restricted account, anonymized: the handle and any quoted posts are hidden.
Shadowban Radar audit of 6 Nov 2026. This is our reading of your account, based on X's official algorithm.
Sources: your 4 tests of 6 Nov at 13:26, the 36 posts and replies you published between 1 Nov and 6 Nov (all of them, read on 6 Nov, 13:59), and your 6 images and videos, looked at one by one.
All times are UTC, because we do not know your time zone.
read in X's published code or pages
read on your account
our deduction
The thing that matters most for your account. This step follows the causes we rank first. The full plan, with its dates, comes further down.
ESTIMATEWrite each reply as one sentence about the post you answer, and spread them over more days.
Covers: your replies and your rhythm. Step 1 of the full plan.
MEASURED. Your last 30 days, read from 1 Nov to 6 Nov.
36
posts and replies read
12 posts and 24 replies, 1 Nov to 6 Nov
25
posts and replies in one day
2 Nov, with up to 12 in a single hour
5
posts with media in 24 hours
the most at once, out of 6 posts with media
67%
of your items are replies to others
24 replies
What we see18 of 24 replies are 5 characters or fewer. The median reply gets 4 views, against about 19 for a post.
ESTIMATEWhat X doesThis shape matches the accounts whose replies we measured as hidden.
What we seeX hides your recent posts from search. The limit touches the way people find your account.
ESTIMATEWhat X doesThe code X published has no visibility rule for search suggestions or search results, so none can be read. A limit on how an account is found fits a limit placed on the account. Reports from other users, which X's models learn from, may feed it, and a public reading cannot see them. This is our reading.
X puts labels on posts and on accounts. We cannot see yours, so we rebuilt them from your public posts, with our reading based on X's official algorithm.None of the post labels we can estimate showed up in your posts.
Your name shows up when typed.
A visible name rules out the cause that hides names from suggestions.
ESTIMATECertainty: low
X publishes no duration for search or reply limits, and we have no date to estimate for your account.
CODEYes. Your pass runs the 4 tests throughout the day, and the score on your dashboard comes from the latest one. We email you within hours when a limit lifts.
MEASUREDNot all at once. No rule we know of lifts a restriction when posts are deleted, and each deletion is one more action that X's behaviour model reads. One exception: when X requires you to remove a post, remove it, because X asks for that before you can post again. Posting and deleting the same content again and again is listed as spam in X's rules.
By 8 NovESTIMATE
ESTIMATEWrite each reply as one sentence about the post you answer, and spread them over more days.
grox/core/lm/thread.py, L48 to L51
Throughout the day
Your 4 tests, until your pass ends. An email within hours when a limit lifts.
Your personal report, practical actions and 14 days of monitoring. One payment. Up to 3 accounts.
Your audit is written after payment, from the last 30 days of your account and its latest checks. It is saved in your Radar account, with its date, and sent by email if enabled. The audit report is written in the language you buy it in.
Check your accountNo subscription. Your pass ends automatically.
| Day | Posts | Replies | Median views |
|---|---|---|---|
| 1 Nov | 1 | 0 | 49 |
| 2 Nov | 6 | 19 | 21 |
| 3 Nov | 3 | 5 | 20 |
| 4 Nov | 1 | 0 | 6 |
| 6 Nov (so far) | 1 | 0 | 2 |
MEASURED. Timestamp of each item, UTC. The median counts your posts, not your replies.
visibility-filtering-client/vf_client.rs, L29 to L44; visibility-filtering/rules/levels.rs, L165 to L476; phoenix/crates/serving/xai-recsys-proto/proto/recsys.proto, L1186 to L1187; agatha/scalding/labels/rate_based_labels/RateBasedLabels.scala, L29 to L57 and L59 to L111
Secondary signal: accounts without a restriction do as much.
What we seeThe burst of 2 Nov: 25 posts and replies in one day, up to 12 in an hour, against about 1 on your other days.
ESTIMATEWhat X doesX's behaviour model reads the rhythm of an account. X publishes no threshold for it.
Certainty: low. It depends on X's image model, which we cannot run. Your visible name in search suggestions also argues against it.
These labels keep posts out of recommendations to non-followers. The code X published gives no rule for search and replies, so we cannot tie these labels to your red tests there. X's help page says that when a post breaks its rules, the limits can include its exclusion from search results and its downranking in replies. We cannot see those limits, so a red test there can come from such a limit.
Not in this estimate: labels set by user reports or by X staff, and any label we found no sign of in your posts.
Your recent posts do not show in search for logged-out readers.
Most likely cause: a limit on how your account is found.
ESTIMATECertainty: low
At least one of your replies is not shown in the conversation we tested.
Most likely cause: very short replies.
ESTIMATECertainty: low
That reply is missing from the ranked view of the conversation.
Same probable cause.
ESTIMATECertainty: low
MEASURED. Score 25 out of 100 on 6 Nov at 13:26, 4 of 4 tests completed. This is the first test on your account.
No. A new account that keeps the same rhythm lands in the same limit, with no history to show that it is you.
ESTIMATEIf the pattern starts again, the restriction restarts too, and X publishes no end date for it. ESTIMATE
Not affiliated with X Corp. X is a trademark of X Corp.