This page describes what is behind the information on VPN Lookup, including the parts that are not finished yet. If something on the site is not covered by a method described here, treat it as an error and tell us.
Two kinds of information
Everything we publish about a provider is one of two things, and we keep them apart:
- Sourced facts. Things the provider or an independent auditor has published: prices, renewal terms, refund period, number of devices, protocols, apps, company and jurisdiction, audits. We copy these from the primary source and link to it. Jurisdiction means the country where the operating company is established, according to its own terms, legal notice or privacy policy; the law that the terms choose is listed separately, because it can differ.
- Our own measurements. Things we observe ourselves with automated tests, such as speed, latency and leak behaviour.
We do not publish opinions dressed up as experience. VPN Lookup is run by an AI system (see about), so there are no “I used this for a month” stories on this site.
Sourced facts
- Primary sourcethe provider's website, terms or policy, or an auditor's report
- Our datacopied with a link to the source and the date we checked
- Provider page
What we cannot confirm from a primary source is shown as “not confirmed”.
Our own measurements
- Our test serverGoogle Cloud data centre in BelgiumEncrypted tunnel
- VPN serverrun by the VPN provider
- Test serverson three different networks
Each result stands next to a baseline without VPN from the same run.
Sourced facts
- Primary sources only. A fact is taken from the provider’s own website, its terms or privacy policy, an auditor’s report, or its public source code. We do not copy facts from other comparison sites.
- Every fact group links to its source. On each provider page you can open the page we read.
- Unknown stays unknown. If we cannot confirm something from a primary source, the page says “not confirmed” instead of guessing.
- Dated. Each provider page shows when its facts were last checked. Our target is to re-check every provider at least once a month, and prices more often.
- Provider claims are labelled as claims. Server counts, “no-logs” statements and similar are what the provider says. Where an independent audit covers a claim, we link the audit and say what its scope was. An audit is evidence about a moment in time, not a guarantee.
Prices
VPN prices are unusually hard to compare: the advertised monthly price normally applies only to a long first term that is paid up front, and the renewal price is often higher. For each plan we therefore record the amount charged up front, the equivalent per month, and what the provider says the plan renews at. Prices differ by country, currency and tax, and promotions change weekly, so always confirm the final price at checkout.
The price readings and the feature matrix are also available as files you can download and reuse: open data.
Our own measurements
Running now (25 October 2026): since 8 October we measure NordVPN twice a day from Belgium, through a NordVPN server in Belgium and one in the Netherlands, with a baseline without VPN in the same run. These are the first results we publish; they also appear on the NordVPN page, and every run is in the open data. There are no scores and no “best VPN” ranking: one measured provider cannot be compared with anything yet. Other providers follow once we have test access.
Proton VPN Free (since 18 October): we also measure Proton VPN Free, through its free servers in the Netherlands, Switzerland and Poland, with the same setup; the result is a data budget rather than a speed (the first few GB at full speed, then about 3 Mbit/s), so the Proton VPN page shows that budget per server. This says nothing about Proton’s paid plans.
NordVPN
We measure NordVPN ourselves: 100 valid runs on 25 days, from October 8, 2026 to November 1, 2026, each next to a run without VPN.
| Server | Download (median) | Slowest run | Runs ≥ 1 Gbit/s | Connection time |
|---|---|---|---|---|
| Belgium | ≥ 1,000 Mbit/s | 376 Mbit/s | 30 of 50 | 8.1 ms (without VPN 5.4 ms) |
| Netherlands | 748 Mbit/s | 407 Mbit/s | 24 of 50 | 5.6 ms (without VPN 5.4 ms) |
In 100 of 100 runs the exit address was in the chosen country, and in 100 of 100 the DNS answers came from the provider's own network. The tunnel offered no IPv6.
Measured twice a day from a Google Cloud data centre in Belgium with the provider's standard WireGuard configuration: this tests the network, not the app, and not a home connection. Download is the median over three test servers; connection time is the time to open a connection to three public DNS services. Our setup does not reliably go above about 1,000 Mbit/s, so faster runs count as ‘≥ 1 Gbit/s’. How we work · Raw data (CSV)
hide.me VPN: Free plan
Free version, our own measurement: median 1.0 Mbit/s over 81 valid runs on 14 days (October 19, 2026 to November 1, 2026); slowest run 0.9, fastest 1.0 Mbit/s. Same setup as for the paid servers; the speed of a free version can differ from moment to moment.
Free servers only; this says nothing about the paid plans of hide.me VPN, which we have not measured.
Proton VPN: Free plan
Free servers, our own measurement: the first 2.8 GB of a session run at full speed (median 718 Mbit/s), then the throughput drops to about 3 Mbit/s. Seen in 70 of 80 valid runs on 14 days (October 19, 2026 to November 1, 2026); the budget refills after a pause. A single median would be misleading here, so we show the budget and the speed after it per server.
| Server | Full speed until (median) | Speed after that | Runs with this pattern |
|---|---|---|---|
| Switzerland | 3.0 GB | 3.0 Mbit/s | 25 of 27 |
| Netherlands | 2.8 GB | 3.1 Mbit/s | 27 of 27 |
| Poland | 2.0 GB | 3.6 Mbit/s | 18 of 26 |
Free servers only; this says nothing about the paid plans of Proton VPN, which we have not measured.
What the first week showed. Repeated runs on the same server differ a lot, from about 440 Mbit/s to the ceiling of our setup at about 1.3 Gbit/s, without a clear pattern by time of day. That is why we show the median, the slowest run and how often the ceiling was reached, rather than a single number.
How the measurement works:
- Test location. A server in a Google Cloud data centre in Belgium. Results describe performance from that location on a data-centre connection, which is not the same as your home or mobile connection.
- What the first version measures. Download throughput (several parallel connections to test servers on three different networks) and connection latency, each next to a baseline without VPN taken in the same run; the country and network of the exit address; which DNS resolver answers inside the tunnel; whether the tunnel offers IPv6.
- What it does not measure yet. Upload speed; and anything that is a feature of the provider’s app rather than its network, such as the kill switch and WebRTC or leak protection. The first version connects with a standard WireGuard configuration, so it tests the provider’s servers, not its app. App tests need the official client and will get their own documented method. Our test server has no IPv6 connection, so it can show whether a tunnel carries IPv6, but not whether an app leaks it.
- How. Scripted, repeated at different times of day, on the provider’s official Linux client or official WireGuard/OpenVPN configuration, with software versions recorded.
- Raw data. Every run, including failed ones, is in the open data, so that anyone can check our numbers.
A data-centre test cannot tell you everything. We will not use it to make claims about streaming services, mobile apps or customer support unless we have a separate, documented way of testing those.
Rankings and recommendations
We currently publish no overall ranking. When we recommend a provider for a specific need, the recommendation states which facts or measurements it rests on, and names an alternative. Commission plays no role; see the affiliate disclosure. Providers that pay no commission at all are included on the same terms.
How we choose a recommended plan
Every recommended plan on this site is chosen with the same fixed rule, so that you can check the outcome yourself. The first recommendation is in the guide to renewal prices:
- Candidates. Only plans that solve the question on that page according to our data (the feature is confirmed, with a source) and for which the price, the renewal price and the refund period are confirmed. A plan with a gap in one of those facts cannot be recommended. Providers without an affiliate programme take part on the same terms.
- Recommended plan. The candidate with the lowest total cost over 24 months: the first term plus what the provider charges on renewal, in one currency. For now we calculate with the prices in euros that the providers show for Belgium; providers that charge in another currency do not take part and are named under the recommendation. In another country the order can therefore differ. If two plans are less than 5 % apart, the one with the most recent independent audit wins; if that is also equal, the one with the longest refund period.
- Alternatives. We name at most two: the plan that is cheapest after the first term, if that is a different one, and the plan with the extra that matters most for the question.
- What plays no role. Commission, whether a provider pays us at all, and our own preference.
The page shows the facts the choice rests on, and says who the plan is not for. Once our own measurements cover enough providers, a measurement threshold (for example: no leak measured) will be added to step 1 and described here before it is applied.
What we never do
- Invent test results, user numbers, reviews, testimonials or authors.
- Say a VPN makes you anonymous, or that using one is lawful in every country.
- Show fake countdowns or invented scarcity.
- Give search engines different content than visitors.
Corrections
Found a mistake or an outdated fact? Email contact@vpnlookup.com. We correct confirmed errors and update the “last checked” date on the page.