Our methodology

How we test, score and rank the VPNs at AvisVPN. Public document, updated quarterly.

1. Our editorial commitment

AvisVPN is an independent site that belongs to no VPN provider. No service can buy a score, buy a recommendation, or pay to alter or remove a negative comment. Our revenue comes exclusively from affiliate commissions paid by VPNs when you subscribe through our links — which changes neither the price you pay nor our opinion of the service.

Concretely, this means we also recommend services like Mullvad or Proton VPN whose affiliate programmes are less lucrative (or non-existent for Mullvad) because they fit certain user profiles. If we wanted to maximise short-term revenue, we probably wouldn't have written this page.

2. How we select the VPNs we test

We only test in depth services that meet the following criteria:

At the time of the last review (April 2026), 8 services meet these criteria: NordVPN, Surfshark, CyberGhost, Proton VPN, ExpressVPN, PIA, Mullvad, PureVPN.

Free VPNs (Hola VPN, Betternet, etc.) are excluded by default with one exception (Proton VPN Free, one of the rare serious free plans on the market, is mentioned in the Proton review).

3. Our five evaluation criteria

Each VPN gets a score out of 10 per criterion, plus a weighted overall score. The five criteria:

CriterionWeightWhat it measures
Speed25%Average throughput loss on a 1 Gbps fibre line
Security25%Encryption, audits, jurisdiction, leaks
Streaming20%Unblocking Netflix, Disney+, BBC iPlayer, etc.
Price15%Real cost over the commitment period
Ease of use15%Install, interface, support

4. Methodology: speed tests

Environment: 1 Gbps fibre connection in France (Paris), Windows 11, Chrome 124, no other active traffic on the network.

Tools: Speedtest.net (Cloudflare and official Speedtest servers), fast.com (Netflix), iperf3 for long-duration measurements.

Protocol: for each destination (France, Germany, UK, US East, US West, Japan, Singapore), we run 5 measurements 30 seconds apart, then average after dropping the highest and lowest values. Tests are spread over 3 different days to neutralise short-term network variation.

Default VPN protocol: WireGuard (or its in-house equivalent: NordLynx for NordVPN, Lightway for ExpressVPN). Other protocols (OpenVPN, IKEv2) are tested occasionally for consistency.

Reference: the "without VPN" throughput is measured just before each session, from the same machine and to the same destination, to calculate a reliable percentage loss.

5. Methodology: security audit

We check seven elements for each VPN:

  1. Default encryption (AES-256-GCM minimum, RSA-2048 or higher for the handshake)
  2. Existence of an independent no-logs audit, recent (under 3 years), with a public report
  3. Jurisdiction of the head office (5/9/14 Eyes vs privacy-friendly jurisdictions)
  4. Leak tests via our DNS test, WebRTC test and IPv6 test, plus dnsleaktest.com and browserleaks.com as references
  5. Kill switch tested by simulating a tunnel drop during a BitTorrent download — the connection must block in under a second
  6. Server architecture: RAM-only (wiped on every reboot) or persistent disk?
  7. Available protocols: WireGuard, OpenVPN UDP/TCP, IKEv2 — which are supported and which are default?

6. Methodology: streaming

We test each VPN against 8 streaming platforms across 3 regions (France, US, UK), over 7 days to measure unblocking stability:

For every platform × region combination, we test 3 different servers. The result is marked ✅ if at least 2 of 3 servers work reliably across the week. Otherwise we mark it as variable or non-functional.

7. Methodology: pricing

We compare prices on three axes:

We factor in bonus months (CyberGhost +4, PureVPN +3, etc.) and included features (password manager, antivirus, dedicated IP) to compute the real value-for-money.

Prices shown are verified monthly and updated if the provider changes them. Seasonal promos (Black Friday, etc.) are flagged in the relevant section.

8. Methodology: ease of use

Score awarded by observing a non-technical user (recruited outside the team, first-time user of the VPN concerned) on:

9. Scoring system

The overall score is not a mathematical average of the sub-scores — it's a weighted score per the grid in section 3, with editorial adjustment for qualitative factors (provider history, recent controversies, support quality).

Scale used:

10. Update cadence

Each review is verified quarterly on the following elements:

The last verification date is shown at the top of each article ("Updated on…"). Major changes are dated in the body when they significantly change the verdict.

11. Sources and references

We systematically use primary sources:

12. Conflicts of interest and financial transparency

AvisVPN is funded by affiliate commissions paid by VPN providers. When you subscribe through our links, the VPN pays us a commission (usually 30-60% of the first month for services that run an affiliate programme, at no extra cost to you).

We clearly mark affiliate links with the rel="sponsored" attribute and a visible label ("Get the deal"). The direct link to the official site without affiliate tracking is always available in our articles.

Transparency on VPNs without affiliation: Mullvad refuses any affiliate programme on principle. We recommend it anyway (its 8.7 score puts it in our top 5) because it's technically excellent. It's a strong editorial signal of our independence.

13. Errors, corrections, contact

We publish corrections when we identify a factual error. If you spot an incorrect data point (outdated price, audit not mentioned, removed feature), email us at contact@avisvpn.com with the source — we correct within 48 hours and note the correction date at the bottom of the article.

Our contact page details other ways to get in touch.

Document updated on 23 April 2026. This methodology evolves regularly — the current version is authoritative.