Human Typing

Reddit removals

“Low-Effort/AI content is auto-removed.”

That sentence, and the line that always follows it, came from a bot in one subreddit — not from Reddit, and not from anybody who read your comment. Here is what the filter could see, why it fires on writing people did themselves, and what to do next.

Low-Effort/AI content is auto-removed. I am a bot, and this action was performed automatically. Please contact the moderators of this subreddit if you have any questions or concerns.
Posted as a reply, seconds after the comment went up.

Where the message comes from

Not from Reddit, and not from anybody who read what you wrote. One subreddit, one line of configuration, matched the moment you pressed comment.

It is the subreddit, not Redditu/AutoModerator
AutoModerator is a bot account every subreddit can configure. Moderators write rules for it in YAML and it applies them as comments arrive. The second half of the message — I am a bot, and this action was performed automatically — is the footer every AutoModerator action carries, which is why the same sentence sits under thousands of unrelated removals.
The reason is a moderator’s own wordingremoval reply
Low-Effort/AI content is auto-removed is a line somebody typed into that subreddit’s config. The capitalisation and the phrasing drift from sub to sub, which is why searching the sentence turns up a scattering of communities rather than a Reddit help page.
Removed is not deletedvisible only to you
The comment still shows on your own profile while being invisible in the thread to everyone else. Most people find out from this reply, or from a comment that never draws a single response.
Nothing about it was reviewedno human in the loop
The removal lands within seconds. Please contact the moderators of this subreddit is not boilerplate — modmail is the only route to a person, and on a false positive it is the route that works.

What a filter like that can actually see

No subreddit publishes its configuration, so nobody outside the mod team can tell you which line matched yours. This is the range of what is available to one — cheapest to write first.

  1. 01

    Account age and karma

    thresholds

    A minimum account age or a karma floor is a single line and catches most of what moderators are trying to catch. If this is what matched, your writing was never examined at all: a comment from a three-day-old account did not get as far as being read.

  2. 02

    The text itself

    regex · length · domains

    Word lists, link domains, formatting, a length ceiling. A long, clean, link-free first comment in a subreddit you have not posted in before is unusual enough to be worth a rule of its own.

  3. 03

    Style tells

    detector bots

    Some subreddits run comments through a third-party AI detector alongside AutoModerator. Those score surface habits — em dashes, tidy parallel structure, an opening that restates the question, the not X, but Y sentence — and they are wrong about careful writers often enough that the false positive you are holding is a common one.

  4. 04

    Repetition

    duplicate text

    The same comment, or nearly the same one, across threads or subreddits. Reposting something that was removed is itself the pattern, and it is what turns one removal into several.

  5. 05

    How the comment arrived

    site-wide filters

    Composition time, and the shape of the input that produced the text. A nine-hundred-word comment that existed in the box for less than a second was pasted, and paste-and-post is the mechanical signature of everything a platform is trying to catch. Worth being exact about where this one lives: AutoModerator never sees your keyboard — it is handed the finished comment and your account. Arrival is measured a layer above it, by Reddit’s own filters and by the detectors some subs bolt on.

Which is why the same comment is removed in one subreddit and stands in the next one over. There is no single detector to reason about, only a stack of them with different inputs and no published rules.

The one input you control end to end

Four of those five are about your account or your words. One is about the mechanics of getting the words into the box — and that is the one you can change without changing what you wrote.

Typing produces a keystroke for every character: uneven gaps, a pause where a sentence ends, a typo backspaced three characters later, a pace that drifts over four minutes. A paste produces one event. The field goes from empty to finished between two frames, and every measurement of how the text got there is a measurement of a machine putting it there.

If you drafted in a notes app, in a doc, or in anything other than the box you post from, that is the shape you hand over every time — regardless of who wrote the words. Human Typing types your draft into the field at a human pace instead, so the trace matches how the writing actually happened.

Said plainly, because it matters more than the pitch: this is not the most likely cause of the message you are reading, and it does not make writing that reads as AI read as human. It removes one signal — the only purely mechanical one on the list, and the one that is misleading about you every single time you draft somewhere else.

What to do with this removal

In order, and the first one is worth doing before the rest.

  1. 1

    Confirm it is actually gone

    Open the thread in a private window. If your comment is not there, it was removed. If it is there, you are looking at downvotes or a blocked user rather than a filter.

  2. 2

    Message the moderators once

    The message names the route itself: modmail. Quote what you posted, say you wrote it, and ask which rule matched. Short and civil gets answered — an argument about the accuracy of AI detection does not.

  3. 3

    Do not repost it

    Sending the same text again is the duplicate pattern, and it is how one automatic removal becomes a ban. Wait for the reply instead.

  4. 4

    Read the subreddit’s rule on AI

    If the sub bans AI-assisted writing and a model wrote it, the removal was correct and nothing on this page changes that. The rule is theirs to set, and it applies to you.

  5. 5

    If it was a threshold, it was never about the comment

    Age and karma floors clear with time and ordinary participation, not with a rewrite. Nothing you could have posted that day would have gone through.

Where Human Typing fits, and where it does not

You arrived here from a search for a removal message, so the honest version comes before the button.

What it does

  • Types the draft you paste into its own panel into the page instead, one key at a time — uneven intervals, pauses at sentence boundaries, the occasional backspaced typo, a pace that drifts the way hands do.
  • Runs entirely inside your browser. No server, no account, no analytics; your text never leaves the machine.
  • Works in ordinary text fields and in the rich text editors that manage their own document model, which is what a modern comment box is.
  • Shows you the estimate before it starts, and stops the moment you tell it to.

What it does not

  • It does not undo a removal, appeal one, or talk to a moderator for you. That is modmail, and it is the step above.
  • It does not touch account age, karma, links or posting rate — the causes most likely behind this particular message.
  • It does not make text a model wrote pass as text you wrote, and it is not a way around a subreddit that bans AI content.
  • Nobody can promise you will never be flagged again. Anything that does is selling you something.