Streaming unblocking is the least stable thing we measure. A provider can go from six-for-six to two-for-six in a fortnight because a platform refreshed its IP blocklist. So this section of our reviews gets re-tested every month rather than every cycle.
How we test
For each provider we pick 36 servers — six per platform, spread across the countries where that platform has a distinct library. We load the service, start a title, and let it play for three minutes. A pass means the video played at full resolution with no proxy error. A partial pass — plays but drops to 480p, or works only on "streaming optimised" servers — is recorded as a fail, because it is not what the marketing promised.
Results
CyberGhost, NordVPN and ExpressVPN all cleared six from six. CyberGhost was the most consistent of the three at the server level: Netflix played on 34 of its 36 sampled servers, against 29 for NordVPN, which matters if you would rather not hunt for a working endpoint every evening.
At the other end, Mullvad failed every platform, and Private Internet Access managed three. Mullvad's result is a deliberate trade-off rather than a failure of engineering — the company does not buy the residential IP ranges that make unblocking work, on privacy grounds, and it says so plainly on its own site. If streaming is why you want a VPN, that is simply not the product.
A VPN that unblocks everything today is not promising anything about next month. Buy on the refund window as much as the test result.
If your stream stops working
- Clear cookies for the platform, then reconnect. Cached location data causes more "proxy detected" errors than actual IP blocks.
- Try a server in a different city in the same country before giving up on the provider.
- Ask support for the current recommended server — the good providers maintain an internal list they will hand over on request.
- Check you are not leaking DNS. A geo-mismatch between your IP and your resolver is an instant block on most platforms.
Why we re-test this monthly
Streaming results decay faster than anything else we measure. A platform refreshing its blocklist can take a provider from six-for-six to two-for-six within a fortnight, and nothing about the provider has changed. Publishing a streaming result annually, as some sites do, is close to publishing nothing.
What a "streaming optimised" server actually is
Most providers maintain a subset of servers on IP ranges that platforms have not yet identified as datacentre addresses, and label them accordingly. That is legitimate, but it means a provider can claim to unblock a platform while only two of its servers actually do. We record that as a fail, which is why our numbers are lower than some published elsewhere.
Working on six servers out of 3,000 is not the same as working. We test ordinary endpoints and report what they did.
Buying advice that survives the next blocklist update
Choose on refund window as much as current results. CyberGhost's 45 days covers a full billing cycle plus change, which means you can confirm the platforms you care about still work before you are committed to anything.
