The videos are still there; the feed just stopped showing them. TikTok's rewritten guidelines, in effect since September 24, limit how far whole categories of content can travel without deleting a single post. When the most important rule works by making things invisible, nobody outside the company can see how it is enforced.

Removed From the Feed, Not the App: TikTok's New Rulebook and the Quiet Power of Who Gets Seen

TikTok's rewritten Community Guidelines took effect Wednesday, and the most consequential rule in them never deletes a single video.

By The Daily Reflection · September 26, 2026 · 9 min read

---

On Wednesday, a new set of rules went live for everyone who posts on TikTok. There was no press conference and no dramatic ban. There was a revised document, and inside it, a line that matters more than any takedown policy: some videos will stay on the app, fully visible to anyone who goes looking, and simply stop being recommended.

Creators already have a phrase for what that means. As one creator-industry outlet put it this week, losing the feed is losing the funnel.

Most of the public argument about social media still runs on an old picture of censorship, where a post is either up or down. TikTok's updated guidelines are a reminder that the picture is out of date. On a platform where almost nobody finds anything by searching, the decision that shapes public conversation is not whether a video exists. It is whether the algorithm will show it to anyone.

Read that again: on TikTok, the most powerful moderation tool is not removal. It is silence.

---

What Changed on September 24

Start with what the rewrite actually says, according to breakdowns published by creator-industry outlets ahead of the effective date.

AI content has to be labeled. Realistic scenes or people created or significantly altered with AI must carry a label. The guidelines put it plainly: "If you're using AI, label it as AI generated content." Unlabeled material can be removed, restricted, or labeled automatically by TikTok's own systems.

Paid promotion has to be disclosed. Anyone promoting a product, brand, or business is expected to switch on TikTok's content disclosure setting before publishing. That lands hardest on the finance creators who have turned the app into a storefront for trading courses, credit cards, and side-hustle pitches.

Certain topics lose reach, not their place on the app. Coverage of the update singles out wellness, finance, and education creators as the groups most exposed to reduced recommendations. The guidelines name patterns such as "dieting and extreme fitness" and "overgeneralized mental health information" as grounds for limiting feed eligibility. Those videos stay on creators' profiles and in search. They just stop traveling.

Conduct off the app can cost you your account. TikTok says it "may also restrict or ban accounts belonging to people who have engaged in egregious off-platform behaviors," pointing to serious crimes and membership in violent or hateful groups.

Enforcement comes in steps. One breakdown of the changes describes an escalating ladder: a warning, then restrictions on features or the account, then a ban. It also reports a new option for some users to complete policy training to restore full access, and says bans can now reach associated accounts, not just the one that broke the rules.

None of this reads as radical on its own. Every major platform labels AI content, polices spam, and bans people tied to violent groups. What makes the update worth a close look is who is writing the rules now, and where the real enforcement happens.

---

The First Rulebook Under New Owners

This is the first major rewrite of TikTok's guidelines since the app's American operation changed hands.

In January, the TikTok USDS Joint Venture took over TikTok's U.S. business. Oracle, Silver Lake, and the Abu Dhabi investment firm MGX each hold 15 percent as managing investors; ByteDance keeps a stake under 20 percent. According to TechCrunch's reporting on the deal, the joint venture oversees content moderation and data protection for American users, while Oracle is responsible for retraining a U.S. version of the recommendation algorithm.

The ownership change was sold to the public as a national security fix. The fear was that a Chinese company could quietly tilt what Americans saw. The solution put a different set of owners in charge of the same machine. It did not change the fact that a small group of people decides what millions of American feeds reward.

Within days of the handover, that tension broke into the open. Users accused the platform of suppressing videos about ICE raids, Jeffrey Epstein, and a shooting in Minneapolis. The hashtag #TikTokCensorship spread on X. California Gov. Gavin Newsom called for an investigation. TikTok said it had not changed its algorithm and blamed a data center outage.

Here is the part most coverage skipped. A team of academics analyzed more than 100,000 videos, NPR reported, and found no evidence of systematic political censorship; posts on every topic had collapsed during the outage, not just political ones. But the researchers also said they could not rule out targeted suppression of smaller numbers of posts, because it would be invisible from the outside. One of them, Benjamin Guinaudeau, summed up the problem: TikTok "can say just about anything" about its algorithm, and "we can't verify it."

That sentence is the right lens for everything in the new guidelines.

---

The Feed Is the Real Rulebook

A removal leaves evidence. The creator gets a notice. The video is gone. There is something to appeal, screenshot, and complain about publicly. Removal is visible power, and visible power can be argued with.

Feed ineligibility works differently. The video is still there. The creator can still share the link. But the view count, which on TikTok depends almost entirely on the For You feed, quietly flattens. Was the video demoted for "overgeneralized mental health information," or was it simply less interesting than yesterday's? From the creator's chair, there is often no way to tell.

This is not a TikTok invention, and it is not automatically sinister. There are real reasons to keep crash diets and pseudo-therapy out of a feed that millions of teenagers scroll every night. A platform that recommends everything equally is not neutral; it just outsources its values to whatever gets the most engagement, and engagement tends to reward the extreme. Choosing not to amplify something is a legitimate editorial call, the same call a newspaper makes about what goes on the front page.

The difference is that a newspaper's front page is public. Everyone can see what it chose. A recommendation system's choices are private, personalized, and unauditable. The categories in TikTok's guidelines are broad enough to cover a certified trainer and a con artist, a licensed therapist and a teenager repeating something they heard. Where the line falls between them is decided case by case, mostly by software. TikTok said last year that over 85 percent of the content it removes is caught by automation, and there is little reason to think feed decisions involve more human judgment than removals do.

When the most important rule is enforced by making things invisible, the enforcement is invisible too.

---

Your Life Off the App Is Now Part of Your Profile

The off-platform clause deserves its own scrutiny, because it quietly widens what a platform claims the right to judge.

The stated targets are hard to defend: people who commit serious violent crimes, members of violent extremist groups. Few users will lose sleep over a terrorist recruiter losing a TikTok account. Other platforms have similar rules for the same reason; a platform that knowingly hosts a violent group's organizer is making a choice too.

The trouble is the machinery around the word "egregious." Who decides what counts, on what evidence, and with what chance to respond? A court convicts on proof. A platform acts on a standard it writes, applies, and reviews itself. Pair that with bans that reportedly extend to associated accounts, and the reach of a single judgment grows: the penalty can follow a person rather than a post.

Now place that power under owners who took control amid a public argument over political bias, in a year when a midterm election is weeks away. The rule has not been shown to be applied unfairly. It is simply the kind of rule whose fairness depends entirely on trust, and trust is exactly what TikTok's new ownership has been struggling to earn since January.

---

The AI Label Problem Nobody Has Solved

The AI labeling rule is the most defensible piece of the update, and also the one most likely to fail quietly.

Labels help. A viewer who knows a clip of a politician or a disaster scene was generated by software can judge it accordingly. TikTok deserves credit for making the rule plain and for promising to label content itself when creators do not.

But the rule leans on two weak points. The first is honesty: the creators most likely to use realistic AI to mislead are the least likely to label it. The second is detection: automatic labeling works only as well as the systems that spot synthetic media, and those systems are locked in an arms race with the tools that generate it. Meanwhile, the clearly labeled, harmless AI content gets the tag, and viewers learn a dangerous lesson: if it isn't labeled, it must be real.

That is not a reason to skip labels. It is a reason not to treat a label policy as a solution to a problem it can only shrink.

---

The Bottom Line

It would be easy to read TikTok's new guidelines as housekeeping: label your AI, disclose your ads, don't join a hate group. Most of it is reasonable, and some of it is overdue.

The deeper story is about where power sits on a modern platform. It no longer sits mainly in the delete button. It sits in the recommendation engine, the quiet system that decides which of millions of daily videos find an audience and which vanish into a profile nobody visits. TikTok's rulebook now openly describes whole categories of speech that will be allowed to exist but not allowed to spread. That may be the right call for diet pills and fake therapy. It is also a template, and templates get reused.

The question for users is not whether TikTok should have rules. It is whether anyone outside the company can check how they are applied. Right now, as the researchers who studied January's censorship scare admitted, the answer is no. The owners changed. The black box did not.

A platform that can make you invisible without telling you why is asking for a great deal of trust. Until outsiders can see inside the feed, trust is all anyone has.

---

The Daily Reflection cuts through the noise to find the stories that actually matter. Follow for thoughtful takes on politics, technology, and whatever's shaping our world.