Method
What this can prove. What it cannot.
X does not publish a “you are shadowbanned” flag. Anyone who claims to read For You ranking weights is lying. FILTERED is a logged-out public-client audit. That is a smaller, truer surface.
Public client
CLEAR
Not ghosted
Public client
WATCH
Friction
Public client
FILTERED
Signal fired
Public client
INCONCLUSIVE
Not enough
We can see
- Whether the account exists, is protected, or is gone.
- Whether a logged-out client can read recent posts (ghost-ban proxy).
- Official reduced-reach wrapping on posts (TweetWithVisibilityResults) and profile interstitials.
- Public view counts on originals versus follower count.
- Reply-heavy diets, off-platform links, and extreme posting volume.
We cannot see
- Search suggestion bans and tweet search bans. X gated typeahead and guest search. Tools that still stamp “search banned: yes” are using dead endpoints or making it up. FILTERED prints N/A on purpose — that row is not a finding about the account.
- Reply deboost (“Show more replies”). It is personalized to the viewer. There is no honest logged-out test without a sterile reference account we do not have.
- For You ranking, ads, or geographic downranking.
How a stamp is chosen
CLEAR means the profile is public, posts load for a guest, no official reach labels fired, and public views are not dead. WATCH means friction without a hard filter. FILTERED means a public signal actually fired. INCONCLUSIVE means we could not measure enough — and we will print that word, because uncertainty is more honest than a fake ban.
Analytics, evidence, compare
Impressions from X Analytics are the only first-party reach number — paste them after the first stamp. The evidence locker is the posts a guest client actually returned. Compare runs two files on the same surface. None of this is a login.
No handle is stored on a server. Recent files live in your browser. The public X web client token used here is the same bearer every logged-out tab already sends.