You post a video. It gets flagged, removed, or shoved into the void with 8 views and no explanation.

Was it the caption? The dancing? That one random word? You appeal, and lose. And just like that, you’re shadowbanned.

According to a now-deleted AMA from a TikTok moderator, you’re not imagining it. The platform’s moderation system is a confusing mix of overzealous AI, stressed human moderators, policy whiplash, and unclear enforcement.

The source for most of what follows is a single Reddit thread. The claims below come from one former moderator’s telling in a Reddit AMA that has since been deleted. TikTok has not published policy confirming them, so where the company does publish figures that test these claims, those are quoted alongside.

TikTok says most of this work is already automated. The company said in August 2025 that moderation technologies, including AI, take down over 85 percent of content removed for breaking its guidelines. Most decisions about your video are made without a person involved.

We broke down a related case in why TikTok flags original content as unoriginal.

Who Actually Reviews Your TikTok Video

Machines make most of the calls, and the human layer is shrinking. People still sit behind appeals and the harder judgment calls, but they see a minority of enforcement decisions and TikTok has been cutting those teams.

If you thought TikTok moderation was just a bunch of bots pulling levers, you’re half right. There are human moderators, and the AMA described them as overwhelmed, under-trained, and reviewing up to 2,000 videos per day.

That figure comes from the AMA alone and TikTok has never published a per-moderator quota, so treat it as one worker’s experience. The direction it points is corroborated by what the company has done since.

TikTok Is Replacing the Human Layer the AMA Described

In August 2025 TikTok said it would restructure its UK trust and safety operations, putting several hundred jobs at risk as it moved to AI-assisted moderation. The same global reorganization hit moderator roles in South and Southeast Asia, notably Malaysia.

A TikTok spokesperson told AFP the company was “continuing a reorganisation that we started last year… concentrating our operations in fewer locations globally”, and that it planned to reshape moderation “with the benefit of technological advancements”.

The union representing affected staff read it differently. John Chadfield, a national officer at the Communication Workers Union, said workers “have long been sounding the alarm over the real-world costs of cutting human moderation teams in favour of hastily developed, immature AI alternatives”.

Speed is the trade TikTok is buying. On its own content moderation page the company states that of the content it removed in 2024, over 98 percent was taken down within 24 hours.

The automated share has climbed too, though not on a like-for-like basis. The same page reports that over 80 percent of the violative videos TikTok removed in 2024 were taken down by automated technology, while its Community Guidelines Enforcement Report for January to March 2026 puts automated removals at more than 96 percent of total removals. That later figure counts AI moderation takedowns for the first time, by TikTok’s own note in the report, so part of the jump reflects a change in what gets counted.

This is also what the “under review” label means in practice. A video sitting under review has usually been flagged by an automated system and is waiting in a queue, rather than being read by a person.

And then there’s the AI.

According to the moderator, most videos are flagged by automated systems long before a person ever sees them. Sometimes those videos never make it to a human because the system thinks they’re fine, or problematic enough to remove outright.

Even when a person does step in, they don’t always have context. They can’t view your account as a whole. They judge that one video in a vacuum, using constantly shifting guidelines. So yes, your video might get flagged for “nudity” just because someone jumped too enthusiastically in a crop top.

What Actually Gets Flagged and Why Context Is Everything

Framing and proportion decide most borderline calls. The same dance, cigarette or child on screen can pass or fail depending on how much of the frame and the runtime it occupies.

Everything in this section comes from the AMA itself.

Twerking? It depends. A short clip in a club is fine. A full video of suggestive dancing with close-up framing? That’ll get flagged, especially if more than 50% of the screen is “butt” or the twerking exceeds 50% of the video’s length.

Filming while driving? Flagged unless you clearly have both hands free and never look at the camera.

Smoking, vaping, or drug references? Even coded language (like “garden party”) can slow down your views or get you pulled from the For You Page.

Women with large chests? They get flagged more, through human bias rather than official policy. The system flags suggestive movement, jumping, and tight clothing more often on larger bodies. TikTok trains against this bias, but it still happens.

Kids on screen without adults? That’s risky. If the system can’t tell the child is supervised or over 13, the video may be flagged, or worse, the account may be labeled unsafe.

And that “educational or entertainment purposes” disclaimer? Meaningless. If you post something controversial, satire or not, it’s the actual content that gets judged, not your intent.

Some Words Can Trigger a Flag Instantly

TikTok has never published a banned-word list, and the AMA’s examples suggest why. The triggers described are contextual and change often, so any list circulating online is a guess at a moving target.

TikTok has an evolving, internal list of words and phrases that can get your video flagged, throttled, or outright removed, even if they’re not being used maliciously.

According to the moderator:

• Saying anything like “rigged” or “stolen” in reference to an election? Auto-flag. No exceptions.

• Using euphemisms for drugs, even clever ones? TikTok’s AI is trained to search for and identify them. “Garden party” and similar phrases are not safe.

• Swear words? Usually fine in music or non-targeted language. But use the same word at someone? That’s bullying.

• Even words like “dolt,” “dense,” or “foolish” have gotten creators flagged, especially when aimed at others in the comments.

• Sarcasm, tone tags like “/j” (joking), or even creative expressions like “Free Luigi” have been flagged simply because the algorithm doesn’t get context.

Worst of all, creators are being punished inconsistently. One person says “Free Palestine” and gets flagged. Another uses a slur and stays up for hours or days unless mass-reported.

What decides it is how the system interprets the word in that moment.

Winning an Appeal Can Still Hurt You

A restored video does not always restore the reach it lost. The AMA describes accounts carrying a penalty after a successful appeal, which is the hardest claim here to verify because TikTok publishes nothing about it.

This one is brutal.

Several creators reported that even if they win an appeal and their content is restored, they still end up facing invisible penalties. Their videos stop reaching people. Their Promote features get locked. Their lives pull 2 viewers instead of 2,000.

The moderator confirmed that repeated appeals, even successful ones, can make the system treat your account as “problematic.” One user even received a message warning that more appeals could result in a permanent ban.

So what do you do? Never appeal? That doesn’t feel right either.

This gray area is why so many creators start to self-censor, having decided that navigating the rules is too risky.

Comment Moderation Is a Whole Other Beast

Comments run on a separate pipeline with separate triggers. The practical consequence is that the category you pick when reporting changes the odds of the comment coming down.

You’d think comment moderation follows the same rules as video moderation, but nope. Different team, different triggers, and often less context.

Comments flagged for bullying or harassment don’t get weighed the same way as video content. A throwaway line like “well that was dumb” could trigger a strike, while someone else saying something clearly hateful gets a pass.

The moderator also confirmed that report category matters. If you report a hateful comment under “hate speech” and it gets ignored, try reporting it under “harassment” or “bullying.” That simple switch can raise the odds of it getting removed.

And here’s the worst part: creators are being held accountable for comment sections too. If a hateful or suggestive thread builds up under your video and goes unchecked, your post, or entire account, can take the hit.

Shadowbans, Throttling, and the Invisible Wall

TikTok does not use the word shadowban, and the AMA says the internal term is throttling. The effect creators describe is a video that stays up while its distribution is capped.

The word “shadowban” gets tossed around a lot, but according to this moderator, it’s real. Only it is called throttling internally.

When you’re throttled, your videos:

• Don’t appear on the For You Page

• Stall at low views (often 0 to 20)

• Can’t be promoted or monetized

• Get skipped over even by your own followers

And it doesn’t take much. A single flagged post can slow down an entire account. Repeat flags, even reversed on appeal, can put your profile in “probation” for weeks, sometimes months. In extreme cases, the account never recovers. It’s permanently classified as risky.

This is why some creators post bangers back to back and still see zero momentum. They’re stuck behind an invisible wall they can’t break through, not without starting over or waiting it out.

Why Some Creators Are Treated Differently

Identical videos get different outcomes because the inputs around them differ. Who reported it, which category they chose, how fast it was spreading and the account’s prior record all feed the decision.

If you’ve ever watched someone blow up for doing the exact same thing you got flagged for, you’re not imagining it. The moderator explained that moderation outcomes depend on several inconsistent factors:

• Who reports the content and what category they choose

• Which moderator sees the content (human bias exists, even with training)

• How fast the video is spreading when it’s flagged

• Whether your account already has prior strikes or “trust issues”

• How recognizable you are (yes, bigger creators sometimes get more leniency)

This explains why accounts impersonating creators can stay up for weeks, while smaller accounts are flagged instantly for a meme or out-of-context soundbite. The same video can draw different outcomes depending on who flagged it and who reviewed it.

And unfortunately, moderators can’t check your account history. Each decision is made based on a single piece of content, in a sea of thousands, with barely any time to assess it.

How to Protect Your Content Without Losing Your Mind

Assume the first decision is automated and plan for it. Keep originals, keep risky experiments off the main account, and weigh an appeal against the reach cost before filing it.

So, what can you do?

Here’s what the moderator (and countless creators) recommend:

• Keep copies of everything. If a video gets flagged, you’ll want the original.

• Upload from a clean device or account if yours has been throttled.

• Avoid risky phrasing, even if it’s sarcastic or joking. The bots don’t get nuance.

• Use third-party tools like Flick or Systeme IO to organize drafts, back up posts, and plan content in a safer space.

• Skip overly risky content on the main account. Use alts for experiments.

• Don’t assume appeals will fix everything. If something gets flagged, weigh whether it’s worth the long-term hit before challenging it.

Ultimately, the goal is to stay visible while still being yourself, and sometimes, that means playing defense even when you’ve done nothing wrong.

TikTok’s Moderation System Is a Mess, and Creators Are Paying the Price

What the AMA described was a system full of gaps, contradictions and inconsistent enforcement. Whether that is simple incompetence or something worse is a separate question.

Good creators are getting flagged. Abusers are slipping through. Appeals are punishing users for pushing back. And nobody really knows what will trigger a takedown until it’s too late.

If TikTok wants to keep creators, and avoid turning into another burnout app, it needs to fix this. That means clarity, consistency, and more transparency about what content is actually allowed.

Until then? Keep backups. Play it smart. And remember: you’re not alone in this mess.

Frequently Asked Questions

Does TikTok use AI or humans to moderate videos?

Both, with machines making most of the calls. TikTok says moderation technologies including AI take down over 85 percent of content removed for breaking its guidelines, and the company has been cutting human moderation teams in the UK and Southeast Asia since 2024.

How fast does TikTok remove a flagged video?

Usually within a day. TikTok states on its transparency pages that of the content it removed in 2024, over 98 percent was taken down within 24 hours, which is why a video can disappear before any person reviews it.

Is a TikTok shadowban real?

The former moderator says the effect is real and that the internal word for it is throttling. TikTok does not use the term shadowban publicly, and it publishes no view threshold that would let a creator confirm one.

Can winning a TikTok appeal still hurt my account?

The AMA says yes, describing accounts that kept a penalty after a successful appeal. TikTok publishes nothing on this, so it stands as an insider claim rather than a documented rule.

Does TikTok have a list of banned words?

Not one it publishes. The moderator describes an evolving internal list where context decides the outcome, so any word list circulating online is a guess at a target that keeps moving.

What does a TikTok community guidelines violation actually mean?

It means TikTok has decided a video breaks one of its published rules, usually after an automated check rather than a human review. TikTok says moderation technology accounts for over 85 percent of the content it removes.

Why do two identical videos get treated differently?

Because the surrounding inputs differ. Who reported the video, which category they chose, how fast it was spreading and whether the account already carries strikes all change the outcome.