How VPNs Actually Perform
Aggregated from tests that visitors chose to share. Not vendor claims, not a paid review - measurements from real connections, with the methodology and the caveats stated in full.
🔍 Run a Test & ContributeLeak Rates by Provider
Share of shared tests where each channel leaked. Lower is better.
We publish a provider only once it has at least 30 shared results. Publishing sooner would mean one person on a bad connection could define a provider’s reputation, which is exactly the kind of number this database exists to replace.
Speed Overhead by Provider
Overhead is the share of your connection speed lost to the VPN, measured against the same connection with the VPN switched off, in the same session. Lower is better. Absolute Mbps figures are shown for context but mean little on their own - they mostly describe the tester’s internet line, not the VPN.
Overhead needs a matched pair: one speed test with the VPN off, then one with it on, in the same browser session. Running both takes about two minutes and is the single most useful thing you can contribute here.
How This Data Is Collected
- Opt-in only. Nothing is recorded unless a visitor ticks the share box after a test. The box is unchecked on every single run and its state is never remembered.
- No IP addresses. We resolve the connection to an ASN, organisation and country, then discard the address. It is never written to the database.
- Provider attribution. Taken from what the visitor selects, falling back to the network operator behind the exit IP. An unrecognised network is recorded as unknown rather than guessed - a wrong attribution would corrupt the averages of a provider that was never involved.
- Minimum sample size. A provider appears only after 30 results.
- Outliers excluded. Physically impossible readings are flagged at write time and left out of every figure above.
Read this before quoting any number here.
Self-selected sample: visitors run these tests when they already suspect a problem, which skews results negative. Figures describe submissions, not a controlled lab benchmark.
VPNMeter is independent and takes no payment for placement. Where we earn affiliate commission it is disclosed on the page in question, and it has no bearing on the figures above - those come from submissions, which we cannot sell.
🧰 Contribute a Result
Every test runs entirely in your browser - nothing is stored unless you choose to share the result.
VPN Results Database - FAQ
Where these numbers come from, what they can tell you, and what they cannot.
Where does this data come from?
From visitors who ran one of our browser-based tests and then chose to share the result. There is no crawler and no vendor feed behind it. Each entry records the outcome, the VPN provider and the exit country - never an IP address.
Is this a fair comparison between providers?
It is honest but not controlled. People reach for a leak test when something already feels wrong, so the sample skews toward problems, and providers with more users generate more submissions. Treat these figures as a signal about how a provider behaves in the wild, not as a lab ranking where every provider faced identical conditions.
Why is my provider not listed?
Either it has fewer than 30 shared results, or the network behind its servers is not one we recognise. We would rather show nothing than publish an average built from a handful of tests, because a number with no weight behind it is worse than no number at all.
What is speed overhead and why not just show Mbps?
Absolute speed mostly measures the tester’s own internet line - a 900 Mbps result says more about their fibre than about their VPN. Overhead compares the same connection with the VPN off and on within one session, which isolates the cost of the VPN itself. It is the only speed figure on this page that is genuinely comparable between providers.
Can a VPN company influence its position here?
They cannot buy one. Placement comes from submitted measurements, and we do not sell it. We do assume attempts at manipulation once these pages start ranking, which is why submissions are rate-limited, outliers are discarded, and providers stay unpublished until they clear the minimum sample size.
Do you store my IP address?
No. The request IP is resolved to a network operator and country and then discarded before anything is written. What remains for abuse prevention is a hash whose salt is regenerated every day, so it cannot be reversed and cannot connect a visitor across two days.