What we mean by censorship

The claim

Apple operates 175 separate App Stores and assigns you to exactly one, by billing address. What that store will let you install is not what another will. There is no App Store on earth where every app is available: even the best-served storefront is missing 12.1% of the widely-used apps we have tested there, and the typical one is missing 17.2%.

We call this censorship. Not because we know why each app is missing — usually nobody outside Apple does — but because the effect is that a person is prevented from installing software on their own device on the basis of where they live, by a mechanism one company built, operates and does not disclose. Article 19 of the Universal Declaration of Human Rights protects the freedom to seek and receive information “regardless of frontiers”. Apple's App Store is organised entirely around frontiers.

An app can be missing because Apple removed it, because a government or court demanded it, because the developer withheld it, or by mistake. Those are different acts with different culprits, and we say which one it was whenever anyone has put it on the record. But all four run through the same machinery, and Apple built the machinery: without a system of per-country catalogues, none of the four could geographically partition anyone's access to anything.

What an open internet looks like

Nothing else on the internet works this way. Registering a domain involves no set of boxes to tick deciding which countries may reach it. Apple does not keep a per-country list of websites your iPhone is permitted to load — Safari will fetch any address you type, in any storefront, and Apple has never suggested it should be otherwise. On a Mac, made by the same company, you can install any software from anywhere.

So the gate is not a property of computers, of phones, or of the law. It is a decision about app distribution, and Apple made it. When the Digital Markets Act required alternative distribution in the European Union, Apple built it within months — which is the clearest evidence available that the gate was always adjustable at will.

Restriction belongs at the payment, not the download

Territorial law does bind some things. Gambling is licensed jurisdiction by jurisdiction; so are some financial services. Apple cannot be expected to take a payment that is unlawful where the payer lives, and we do not ask it to.

But a payment is not a download. The regulated industry Apple's argument leans on hardest has already demonstrated the alternative: sports betting is licensed state by state in the United States, Apple's storefronts have no concept of a state, so the apps are installable across the whole country and verify location at the moment of the wager. It works, at scale, in the most heavily regulated category there is, inside Apple's own store. Streaming services do the same with territorial licences — Spotify's catalogue differs by country and the Spotify app is installable nearly everywhere. The licence constrains what is inside the app, not whether the app exists for you.

And the commerce argument barely touches most of what is missing. Of the 143,074 apps we hold, 94.6% are free to download — no payment, no tax, no licence for territorial commerce law to attach to. If the partition were tracking commerce, the 6,865 paid apps would be the ones withheld. They are not: 37.4% of paid apps are missing from at least one storefront, against 40.4% of free ones. Free apps fare slightly worse than paid ones, which is the opposite of what a rule about money predicts. Whatever the regional partition is doing, it is not tracking whether money changes hands. (Apple reports nothing about in-app purchases, so “free” here means free to install, exactly as the App Store means it.)

The part that is not arguable

Apple publishes no record of app availability. There is no list, no API, and no page in the App Store that tells you what it is not showing you. Our record holds 263,537 storefront readings where an app could not be got, and Apple has published an account of none of them.

Reasons are a little different, and this is where we are careful. Apple has explained individual removals when the pressure was high enough, and we publish those explanations where they exist — we can currently cite a documented reason for 12 apps. In most of those cases Apple explained privately, to one developer, and the public knows only because that developer chose to speak.

Apple does publish something, and it is worth being exact about what. Its App Store Transparency Report gives the number of apps it removed on a government's demand — 2,045 of the 166,899 it removed during 2025 — and the data file issued alongside it names the government body and the specific statute behind every one of those 2,045: Roskomnadzor under Article 15.1 of Russia's information law for 1,209 of them, the Cyberspace Administration of China under its rules on services capable of social mobilisation for 117, Vietnam's broadcasting authority under Decree 147 for 332. It names no app, in any year. Both halves of every answer are in the same release, and putting them together is the one thing Apple declines to do. Until 2021 it also reported how many removal requests it received and how many it refused, which is the only way to tell compliance from initiative; that series was discontinued while every other part of the report carried on. Apple's 2025 App Store Transparency Report

That gap is why this project exists. We test Apple's catalogue servers for every app in every storefront, publish every reading with its date, and let anyone re-run any test from their own browser and check our answer against Apple's.

What we cannot show

Our readings show that an app cannot be got in a storefront and that it is still available in others. They do not show intent. A developer's market decision and a government's demand look identical from outside, which is why no cause appears anywhere on this site without a citation. Two further limits are worth knowing: roughly four in five apps recorded as unavailable were never witnessed available in that storefront, because monitoring began in February 2019 and anything removed before then is invisible to us; and our headline measure selects apps by worldwide user ratings, which cannot see the apps censorship targets most precisely — the largest Tibetan app we hold has 749 ratings and the largest Uyghur app 1,037.

If you think we are wrong

The definition is worth arguing about and we would rather have the argument than avoid it. The facts are not up for argument: if a page here states or implies something you can correct — a removal that was your own decision, a reason we have missed, a reading that is wrong — tell us and we will publish the correction.

How the testing works →Every storefront, ranked →Contact →