First, measure it properly
Most “my VPN is slow” reports are not measured at all, and half the rest compare a VPN result against a reviewer’s number from a different country. The only meaningful test is the same test, on the same connection, minutes apart:
- Disconnect the VPN. Run the speed test.
- Connect the VPN. Run it again.
- Divide. The second figure as a percentage of the first is your overhead.
A good provider on WireGuard to a nearby server keeps 80–95%. If you are seeing that, the VPN is working correctly and the slowness is somewhere else — your line, the site you are loading, or the time of day.
1. The protocol
This is the largest single factor and the easiest to change. OpenVPN can halve the throughput of a fast connection; WireGuard-family protocols barely touch it.
Open the app’s settings and set the protocol to WireGuard, NordLynx or Lightway. If it is set to “automatic”, it has usually already chosen one of those, but on restricted networks it may have fallen back to OpenVPN without telling you. Set it manually and re-measure.
The exception is a censored country, where the obfuscated modes are slower by design and there is no alternative. See VPN protocols explained.
2. The server
Distance costs latency, and on a single connection latency limits throughput regardless of bandwidth. A server 200 km away and one 8,000 km away can differ by a factor of five on the same provider.
Choose the nearest city that does the job. If you are changing region for streaming, choose the nearest server in that region rather than the first one in the list. If the app shows server load, prefer one under 50% — a full server shares its bandwidth among everyone connected.
3. The device
A VPN is CPU-bound on the client. An older phone or a low-power laptop can be the bottleneck before the network is. Two specific cases worth checking:
- On a Mac, make sure the app is a native Apple silicon build rather than running under Rosetta.
- On a router, the router’s processor decides everything; a budget router on OpenVPN can reduce a gigabit line to tens of megabits. See the router guide.
4. Antivirus and security software
Products that inspect network traffic add their own overhead inside the tunnel, and some interact badly with VPN drivers. Temporarily disable the antivirus’s web-shield or network module, re-measure, and turn it back on. If that was the cause, exclude the VPN adapter in the antivirus settings rather than leaving protection off.
5. The connection itself
Wi-Fi on a congested band, a long powerline run, a mobile connection at peak hour: none of these are the VPN’s fault, but the VPN gets blamed because it is the thing that changed. Test on Ethernet if you can, or right next to the router, before concluding anything.
What is not worth doing
Turning off encryption — no reputable provider offers it, and it would defeat the point. Buying a “faster” VPN before checking the protocol and the server; the difference between providers on a like-for-like test is much smaller than the difference between OpenVPN and WireGuard on the same provider. And paying for a higher tier: none of the providers on this site charge more for speed.
What we can and cannot tell you
Every speed claim in the reviews on this site is currently marked as pending, because we have not yet run our own throughput tests on a stated line with dates attached. Published third-party numbers vary so widely by test location that quoting them as fact would be dishonest. The speed dimension is excluded from each provider’s score rather than guessed, and the speed test here measures your connection, which is the only number that applies to you.