How we measure VPN and proxy availability in Russia
Freedom Checker does not aggregate anyone else's data and does not ask services how they are doing. Every number on the site comes from our own measurements, taken on physical hardware that sits inside Russia and is connected to the same carriers the reader uses.
Where the measurements come from
Checks run from physical probes in three macroregions: European Russia, Siberia and the Far East. Each probe is connected to several carriers at once — one wired and several mobile — and switches between them, running the same set of checks through each.
| Regions | European Russia, Siberia, the Far East |
|---|---|
| Carriers | Rostelecom and Sibirsky Medved (home broadband), MegaFon, MTS, Beeline (mobile) |
| Mobile links | A separate eSIM profile per carrier on the probe's modem — traffic goes over the real cellular network, not through a tunnel to it |
| Cycle | A full round starts every four hours. In practice one node × service × carrier combination is re-checked every 4–6 hours (median 4.6 h across 165,000 measurements over two days); each row shows the time of its last successful check |
This is the fundamental difference from testing “from a hosting provider”. Censorship in Russia works such that the path to a site from a data centre and the path from a mobile subscriber are different paths with different outcomes. You have to measure from where the reader actually sits.
What exactly is checked
For every carrier × service combination the probe brings up a connection and tries to open a set of target resources through it: Instagram, YouTube, Telegram, WhatsApp and a site outside the allowlists. What is checked is not whether the connection was established, but whether real content comes back through it — a tunnel can come up and still carry nothing.
What “working” means
- Working — the connection came up and the target resource returned the expected content through that carrier in that region.
- Not working — the connection failed, or came up but the resource behind it was unreachable.
- In question — the result is unstable between rounds. Such a record is not shown as working: it goes back for a re-check, and the verdict is made later.
That middle state is not decoration. Availability in Russia changes over the course of a day, and a single failed response does not yet mean a block — just as a single successful one does not mean stability.
How a service gets into the checker
We buy the subscription with our own money and use it covertly for several months before adding the service to the site. No service knows it is being measured until it appears in the table.
The point is that a service which knows it is being tested can behave differently: assign a dedicated server, raise the priority, avoid blocking one particular account. A covert test running for months shows what an ordinary paying user gets, not a showroom.
The hardware — modems, SIM cards, probes, traffic — is bought and paid for by us. Nobody pays to be included in the checker, and no such option exists.
About affiliate links — plainly
Some services approach us after they appear in the checker and offer an affiliate link. We do place such links, and that has to be said out loud. Three facts come with them:
- the link appears after the service has already passed testing and entered the table — it was never the reason it got there;
- the amounts are token and go back into hardware and connectivity; the project makes no profit;
- the link has no effect on the measurement — everything is measured identically, and a service with an affiliate link will show a red cross just the same once it stops working.
What we do not measure
- Speed and latency. We check whether something is reachable, not how good the connection is.
- The security and privacy of a service. We do not audit what a provider does with your traffic and logs. Whether something works and whether it can be trusted are separate questions.
- Every carrier and every city. Five carriers and three macroregions are not all of Russia. Your neighbourhood provider may show a different picture.
- The future. A measurement describes a moment. Blocking changes, and the table shows a check that has already happened, not a guarantee for tomorrow.
The data
The current snapshot is published machine-readably: snapshot-lite.json for the default region, snapshot.json for every region in full. JSON, open access.
The data is released under CC BY 4.0: you may freely use it, republish it and build on it, including commercially, as long as you credit the source with a link to freedomchecker.ateo.digital.
Errors
The method is not infallible. A probe can lose connectivity, a carrier can hand out an atypical route for a while, a service can change its configuration between rounds. If you see a number on the site that contradicts your own experience, write to us — @Ateo_help_bot. Most often a discrepancy means you are on a different carrier or in a different region, but sometimes it means a probe of ours has broken, and that is worth knowing.