A content moderation policy guide is only useful once you find the version that applied

A content moderation policy guide starts with a simple fact: a service's rules are not one document. They are a stack of them, written at different times by different teams for different audiences, and the one people quote is usually the shortest and least operative.

This page explains how to read that stack and find an older version. The named policies below are examples of where to look, not statements about what any service currently permits; check the service's own current pages for that.

What to take away

The layers, and what each one decides

The terms of service or user agreement. The contract. It decides what the service may do to your account and your material, how the terms change, and how the relationship ends. It is the only layer with contractual force and the one least often read. What it typically contains, and the asymmetries built into it, are set out in the shape of the platform relationship.

Community guidelines or content policy. The public statement of what is not allowed. Written for readers, deliberately general, and the one everyone quotes. It tells you the categories; it rarely tells you where the line inside a category sits.

Product and feature policies. Separate rule sets for advertising, commerce, monetization, live video, direct messaging, or developer access. These are frequently stricter than the general policy and are the layer people are surprised by, because nobody looks for a second rulebook.

Help center articles. Operational description: how reporting works, what an appeal is, what the account states are. Not rules, and often the clearest writing on the site.

Enforcement guidelines. The operational translation reviewers actually apply, with definitions and worked examples. Usually not published. The gap between this and the public rule is where most arguments about consistency actually live, which is the subject of how a moderation system is built.

Named policies and app-store rules

Start with the platform whose decision you are examining: Meta's Community Standards, YouTube's Community Guidelines, X's Rules, or TikTok's Community Guidelines. Check the general rules alongside any policy for the specific surface involved, such as advertising or commerce, and the help-center articles explaining appeals and enforcement. YouTube's published Community Guidelines enforcement material, for example, explains its warning and strike system; that is not the same as an internal reviewer guideline.

App distribution has a separate policy layer. Apple's App Store Review Guidelines and Google Play Developer Program Policies govern app review and store availability; they are not, by themselves, the rules a social platform applies to an individual post.

Reading order for a real question

What the words are doing

Policy documents use a small vocabulary in a specific way, and misreading it is the usual cause of a surprise.

A rule that looks like a business decision may be the service's implementation of something imposed on it. That implementation may be broader than the underlying obligation or applied across markets rather than only where required.

The routes are set out in how a rule reaches you through somebody else.

Why the same words mean different things in different places

Documents in this area borrow terminology from legal systems that do not agree with each other. The concepts genuinely differ, and reading a policy through the wrong system's assumptions produces confident errors.

Two examples of how much sits under a familiar word. The relationship between a service and the material its users post is treated differently in different systems.

In the United States, Section 230 of the Communications Decency Act (47 U.S.C. § 230) addresses, among other things, platform liability for user-posted content. The European Commission's Digital Services Act page describes the EU's rules for online services; the Act is Regulation (EU) 2022/2065.

Those are two systems, they are not the only two, and neither describes what applies to you. They show that a policy written to satisfy several regimes at once will contain provisions that make no sense from inside any single one of them.

Finding the version that applied

Policy pages are revised, frequently without a changelog, and the previous version often stops appearing on the service's site. If a decision was made about something you published in March, the rule in force in March is the relevant one. A dated copy can help establish what it said, but first check whether an official update notice gives the effective date.

Dated policy changes show why announcement dates and effective dates matter. YouTube announced an update to its hate speech policy on June 5, 2019. TikTok announced revised Community Guidelines on April 17, 2024, with those guidelines taking effect May 17, 2024. For either example, an archived capture before and after the stated change can help show what the public page said at each point.

Take a copy when you make a decision that relies on the rule, not when you need it. Save the page with its date, or save an archived capture with its timestamp and address. If you are checking a past decision, use the retrieval steps above rather than treating a current page as proof of an older version. The general handling of a rule that changes silently is in recording a change to your own record.

What reading the rules will not tell you

The published rule tells you what the service says it prohibits. It does not predict what will happen, because enforcement is automated, sampled and adjusted without announcement, and because thresholds change without any document changing. Treat the published policy as the standard an appeal will be assessed against rather than as a forecast.

It also tells you nothing about what happens once an action has been taken against you, which is a different subject with its own sequence: that is covered in the outcomes and what follows each one.

Common questions

Which document wins if two contradict each other?

The agreement usually says, and it usually says the service decides. In practice the more specific product policy governs the surface it covers. If the contradiction matters to you, the honest answer is that this is a question for the service's own support channel, in writing, before you rely on either reading.

Is it worth reading all of this before signing up?

Read the parts that decide something: the license you grant over what you upload, how the terms change, how the agreement ends, and what happens to your material when it does. The rest can wait until it becomes relevant.

The policy does not mention what I want to do. Does that mean it is allowed?

Not reliably. Enforcement provisions are typically drafted broadly enough to cover conduct the specific rules do not describe, so the absence of a prohibition is weaker evidence than people assume.

Why are the rules so vague?

Partly because precision invites gaming, and partly because a public rule has to be short enough to read and general enough to cover situations nobody has thought of. The precision exists; it lives in the unpublished guideline.