Deletions

Posts recorded as deleted from public Telegram channels

Which channels had posts we held go missing from their public page — recorded only after two separate readings, and reported as a count, never as a reason.

691,351 posts across 83,587 channels are recorded as no longer on the public page they were published on. The median time before such a post was seen missing was 29 hours. The boards below cover only posts published in the last 30 days, on channels that clear the gates.

These boards record detection, not explanation. They count posts the register held that are no longer on a channel’s public page, and when. They cannot say why: an expired ad placement, a repost, a rotating link and a correction all leave the same gap. A share is a share of the posts we hold that were published in the last 30 days, not of everything a channel has ever posted. Neither board is a ranking of misconduct, and a channel on one is not thereby shown to have done anything. Channels below the gates — fewer than 10,000 subscribers, fewer than 5 recorded deletions, or fewer than 30 held posts in the window — are not shown. These are recorded detections: they miss deletions we could not see and may include a small number of wrong records, so no count here is an exact total. The count board also reflects how much a channel posts and how often we read it. A channel we read less often shows fewer.

Most posts recorded as deleted

Channels ordered by how many of their posts published in the last 30 days we recorded as deleted. A count follows how much a channel publishes as well as what happens to it, which is why the share of held posts is printed beside every count here, and the count beside every share on the other board.

Largest share of posts recorded as deleted

The same channels ordered by the share of their held posts, published in the last 30 days, that we recorded as deleted. The gates are what stop a channel with a handful of posts from reaching the top on a single removal: at least 5 recorded deletions among at least 30 held posts. Each board shows at most 100 rows, and there is no paging.

How a deletion is recorded, and what the boards cannot see

Two sightings before a post is recorded. A post is recorded as deleted only after two reads, taken from different exit IPs at least an hour apart, have each found it missing from the channel’s public page. One absence is never a deletion: a throttled, truncated or blocked page looks exactly like a removed post.

What a read is allowed to assert, and what cannot be seen. Any read in which more than 5 of the posts we hold are missing, or more than 3 are missing and they are over a third of the posts we hold in the window, asserts nothing: the read itself is the thing in doubt, and it is refused whole rather than trimmed. The last-24-hour counts of those refusals are in the table above. Only a post we held and later re-read inside a page window can be seen to go, so a post deleted before we first read it is invisible, and a channel we read less often shows fewer deletions than one we read often. These are recorded detections: they miss deletions we could not see and may include a small number of wrong records, so none of them is an exact total.

What the time columns mean. The time is measured to when we first saw the post gone, not to when it was removed, so it is an upper bound on how long the post was up. “Most recent seen missing” is the latest such sighting on that channel.

A post withheld from a country can look the same. Telegram can withhold a post from readers in a particular country. A post withheld from the country our reader exits through is, from here, indistinguishable from a deleted one.

Recorded deletions held out, not confirmed. 7,162 recorded deletions are excluded from these boards, because a check on 3 November 2026 found wrong records concentrated in one place on the page, and these could not be confirmed by a re-check. We stopped recording deletions there the same day and set aside for a re-check the 42,596 already recorded in or near it. Any found present on two separate connections were retracted. Those the re-check could not settle stay out of every deletion count here. The re-check has finished: 35,001 of those recorded deletions were confirmed gone, 433 were found still up and retracted, and 7,162 could not be settled (739 of them because the register no longer holds a current primary handle for the channel). The methodology sets out the check, and what it can and cannot tell us.

Other ways in

The measurement behind the boards, in full, in the methodology. Subscriber movement is on risers and fallers, and the whole register is arranged by measured subscriber size and the language a channel publishes in. Or search a handle directly.