How we score VPNs
How every VPN on this site gets its score: six sub-scores from published facts, fixed weights, the criteria and points behind each one, and the rules for missing data, red flags and ties.
The score in one paragraph
Every VPN gets six sub-scores from 0 to 10, built only from facts the provider publishes: audits, policies, apps, network and prices. The overall score is their weighted sum, also from 0 to 10. A red flag caps it at 6.0. The main ranking lists 377 consumer VPNs with enough published data; the other 455 services are scored the same way and shown with a “low data transparency” note or, in catalogue segments, listed without a score.
Six sub-scores and their weights
| Sub-score | Weight | Criteria | Points for a full score |
|---|---|---|---|
| Privacy and jurisdiction | 30% | 12 | 15 |
| Security and protocols | 20% | 12 | 14 |
| Price and fair terms | 15% | 10 | 12 |
| Censorship bypass | 15% | 12 | 10.5 |
| Network and streaming | 10% | 8 | 8.5 |
| Platforms and support | 10% | 10 | 10.5 |
Privacy and jurisdiction (30%)
Points add up and are divided by 15, the points needed for a full 10; the result is capped between 0 and 10.
| Criterion | Points |
|---|---|
| No logs policy | +2 |
| Independent no-logs audit | +3 |
| No logs confirmed in court | +1 |
| Policy says connection IPs are not stored | +1.5 |
| Policy says session times are not stored | +1 |
| Jurisdiction and surveillance alliances: outside 14 Eyes 1, 14 Eyes 0.4, 9 Eyes 0.3, 5 Eyes 0, not stated 0 | +2 |
| Warrant canary | +1 |
| Transparency report | +1 |
| Sign-up without email | +1.5 |
| Account number sign-up | +0.5 |
| Stores connection IP (penalty) | −3 |
| Public incident (penalty) | −2 |
Security and protocols (20%)
Points add up and are divided by 14, the points needed for a full 10; the result is capped between 0 and 10.
| Criterion | Points |
|---|---|
| Kill switch | +2 |
| DNS leak protection | +1.5 |
| WireGuard or an own modern protocol | +1.5 |
| RAM-only servers | +1 |
| Post-quantum encryption | +1 |
| Perfect forward secrecy | +0.5 |
| Security audit | +2 |
| Open-source clients | +1.5 |
| 2FA | +1 |
| Multi-hop | +1 |
| Bug bounty | +1 |
| Only PPTP / L2TP (penalty) | −2 |
Price and fair terms (15%)
Points add up and are divided by 12, the points needed for a full 10; the result is capped between 0 and 10.
| Criterion | Points |
|---|---|
| Cheapest price per month: percentile among consumer VPNs, cheaper scores higher; no published price scores 0 | +2.5 |
| Free plan | +1.5 |
| Free trial | +1 |
| Trial without a card | +1 |
| Refund window, in proportion up to 30 days | +1.5 |
| Devices at once, in proportion up to 10 | +1 |
| Crypto payments | +1 |
| Renewal price jump (penalty): renewal price per month over the cheapest 12+ month plan, from 1.3× up to the full penalty at 3× | −2 |
| Cash payments | +0.5 |
| Fair terms, each: refunds on monthly plans, pro-rata refunds, auto-renewal off by default, cancelling without support, a renewal reminder, plans shorter than a month | +0.5 each |
Censorship bypass (15%)
Points add up and are divided by 10.5, the points needed for a full 10; the result is capped between 0 and 10.
| Criterion | Points |
|---|---|
| Stealth / obfuscation | +2 |
| VLESS / Reality | +1.5 |
| Trojan | +1 |
| Hysteria2 | +1 |
| Shadowsocks | +1 |
| AmneziaWG | +1 |
| obfs4 | +0.5 |
| TUIC | +0.5 |
| NaiveProxy | +0.5 |
| Mirrors or Telegram channel | +0.5 |
| VoIP in the UAE (claimed) | +0.5 |
| Manual configs | +0.5 |
Network and streaming (10%)
Points add up and are divided by 8.5, the points needed for a full 10; the result is capped between 0 and 10.
| Criterion | Points |
|---|---|
| Number of server countries: percentile among consumer VPNs (log scale) | +2 |
| Number of servers: percentile among consumer VPNs (log scale) | +1.5 |
| 10 Gbps servers | +1 |
| Dedicated IP | +0.5 |
| Port forwarding | +0.5 |
| MENA locations | +0.5 |
| Streaming claims: Netflix, BBC iPlayer, Disney+, Prime Video; two or more gives the full points | +1.5 |
| Torrenting allowed | +1 |
Platforms and support (10%)
Points add up and are divided by 10.5, the points needed for a full 10; the result is capped between 0 and 10.
| Criterion | Points |
|---|---|
| Apps for Windows, macOS, Linux, iOS and Android, in proportion of the five | +3 |
| Routers | +1 |
| An app for Android TV, Apple TV or Fire TV | +0.5 |
| Browser extensions | +0.5 |
| Split tunneling | +1 |
| Ad / tracker blocker | +1 |
| 24/7 live chat | +2 |
| live chat | +0.5 |
| Support on Telegram | +0.5 |
| Status page | +0.5 |
Rules
Normalisation
Each sub-score is the sum of its points divided by the points needed for a full score, capped between 0 and 1 and shown out of 10. Counts such as countries and servers are compared as percentiles among the 544 consumer VPNs in the source data, on a log scale, so a provider with 3,000 servers does not drown everyone else. Price is a percentile too: cheaper scores higher.
“Not disclosed” scores zero
A feature counts only when the provider states it. If a policy does not say whether something is collected or offered, it earns no points, which is not the same as a penalty. Penalties apply only to facts the provider states: keeping connection IPs, session times, a public security incident, PPTP or L2TP only, a large renewal price jump.
Claimed vs confirmed
Claims and confirmations are separate criteria: a no-logs policy earns 2 points, an independent no-logs audit 3 more, a no-logs claim confirmed in court 1 more, a security audit 2. Data a provider sends us counts only after we have checked it against what it publishes.
Red flags and the 6.0 cap
A VPN that stores connection IP addresses, had a public security incident or is under Chinese jurisdiction cannot score above 6.0, however strong the rest is. The reason is shown on its page.
Ties
When two VPNs have the same score, the one with more published data ranks higher, then the one with the smaller renewal price jump.
Weight presets on topic pages
Some rankings weigh the sub-score that matters for that topic more. The chosen sub-score gets the preset weight and the other five share the rest in their usual proportions. Red flags still apply. The weights used are listed below and on each page.
Who is in the main ranking
Only consumer VPNs (six segments: consumer, free and mobile apps, VPNs in security suites, gaming VPNs, decentralised VPNs, Smart DNS) with at least 60% of the data points published: 377 services today. Business, self-hosted, router, provider, browser and white-label services have their own catalogue pages.
Weight presets
Order of columns: privacy, security, price, censorship bypass, network, platforms.
| Preset | Used on | Privacy | Security | Price | Censorship | Network | Platforms |
|---|---|---|---|---|---|---|---|
| Led by privacy and jurisdiction | Anonymity · Jurisdictions | 45% | 16% | 12% | 12% | 8% | 8% |
| Led by security and protocols | Security | 22% | 40% | 11% | 11% | 8% | 8% |
| Led by price and fair terms | Price and payment | 21% | 14% | 40% | 11% | 7% | 7% |
| Led by censorship bypass | Protocols · Travel and restrictions | 21% | 14% | 11% | 40% | 7% | 7% |
| Led by network and streaming | Use cases · Servers by country | 20% | 13% | 10% | 10% | 40% | 7% |
| Led by ecosystem | Platforms · Support · Languages | 22% | 14% | 11% | 11% | 7% | 35% |
| Base weights | Main ranking · Segments (except business) and other | 30% | 20% | 15% | 15% | 10% | 10% |
What we do not measure, and why
No speed tests
A speed test measures one server, from one place, at one moment, over one internet provider. The result says more about the route than about the VPN, and repeating it fairly for hundreds of services is not possible. We score what stays true for everyone: protocols, server network, published terms.
No blocking tests
Whether a VPN works in a restricted country changes week to week and differs by city and mobile operator. Where we mention streaming or access in restricted countries, it is the provider’s own claim and labelled as such.
User reviews do not change the score
Ratings on Trustpilot, the App Store and Google Play are easy to influence, depend on the app more than on the service, and mix very different expectations. We show them on each service page, smoothed so that a handful of reviews cannot outrank thousands, but they never change the score, directly or indirectly.