How we work: testing, scoring and editorial rules

By VPN Clue Editorial Team Updated

VPN Clue compares VPN services using facts from the providers’ own pages. This page explains how we check those facts and how we turn them into a score. It also covers who decides what we publish, how the site earns money and what happens when we get something wrong.

How we test

One record for each provider

Every provider we cover has a structured record with more than sixty fields. The record covers these subjects:

  • plans, prices and refund terms
  • the server network
  • protocols and security features
  • the logging policy
  • independent audits and documented incidents
  • streaming claims, apps and support options

Each section of the record stores the date we last verified it and the official page we checked. Every table and chart on our reviews, comparisons and deal pages is built from this data. That means a table and a chart can never disagree about a price. The text around them can fall behind a change, so a daily check flags any page that quotes a price or count that no longer matches.

Anyone can download the whole dataset as JSON or CSV from our open data page.

What “verified” means

A fact counts as verified when we have read it on the provider’s own pages:

  • prices on the pricing or checkout page
  • the money-back window in the refund policy
  • locations on the server page
  • protocols and app features on the feature or help pages
  • logging rules in the privacy policy
  • audits in the provider’s announcement or the auditor’s report

We never copy a number from another review site and we never fill a gap from memory. When an official page does not state a value, our record says that it is not stated.

Prices are in US dollars as a visitor in the United States sees them. For each plan we note the effective monthly price, the total you pay up front and the renewal price when the page states one. A scheduled job reloads the official pages every day. Each change goes into the price history and the list of recent changes on the open data page.

What we do not claim yet

We publish no speed results for any provider. A single speed test from one laptop on one day says very little. The speed part of our score stays empty until our own lab can measure it night after night. Until then every review shows a “not yet tested” notice and the score uses the other dimensions only.

We also never say that a VPN works with a streaming service unless we have tested it. When a provider lists a service on its own pages, we record that as the provider’s claim and note the date we read it.

The test lab we are building

Our next step is an automated lab. A few servers in different regions will connect to each provider every night through the official Linux apps or standard WireGuard and OpenVPN settings. They will measure:

  • throughput, the amount of data that gets through each second
  • latency, the delay before data arrives
  • connection time
  • DNS, WebRTC and IPv6 leaks

Each published result will be the middle value of the last 30 nights, known as the median. The raw data will be free to download, and we will publish the full method before the first result. Speed joins the score only when a provider has at least 30 days of data.

Audits and incidents

For each independent audit we list the auditor, the date, what it covered and a link to the report or the provider’s announcement.

We also record incidents such as server seizures, breaches, logging disputes and court cases. Each one gets a date, a short factual summary, a severity level and a source. A medium or high severity incident lowers the privacy score for five years. A low severity event is listed without a penalty, for example a server seizure that exposed no customer data.

How we keep pages current

Each page shows its publish date until its content changes. After that it shows the date of the latest change, and its changelog says what changed and when. We keep a separate note of when we last checked each page. A check that finds nothing new should not make a page look newer.

Every week we re-check the pages that have gone longest without a check. If nothing changed, only that note moves. Law and age-verification pages get a check at least every 30 days.

How we score

Every provider gets a score from 0 to 10 in eight dimensions. The total is the weighted average of those scores. The table below is built from the same file the scoring code uses, so it always shows the weights in force.

DimensionWeightInputs
Privacy and jurisdiction 25% Jurisdiction and intelligence-sharing alliance, independent no-logs audit and its age, RAM-only servers, no-logs policy and anonymous payment options. Medium- and high-severity incidents in the last five years deduct points.
Security features 15% Protocols (WireGuard, OpenVPN, IKEv2, proprietary), kill switch on desktop and mobile, obfuscation, multi-hop, post-quantum encryption, own DNS resolvers, DNS leak protection, IPv6 handling, audited apps or infrastructure.
Speed and reliability 20% Throughput, latency and connection time measured by our own lab. Not yet available: this dimension is excluded and the remaining weights are renormalised until the lab runs.
Streaming and unblocking 10% Streaming services the provider documents support for (counted as claims until we test them) and smart DNS availability.
Apps and usability 10% Platform coverage, simultaneous device limit, router support, browser extensions, open-source clients, split tunneling.
Price and value 10% Lowest effective monthly price, monthly-plan price, money-back guarantee length, free tier, renewal pricing honesty, payment flexibility.
Support 5% Live chat availability, email support, knowledge base. Measured response times once we have them.
Transparency 5% Transparency reports, warrant canary, open-source clients, number of independent audits.

Privacy and jurisdiction

A provider earns privacy points for each of these:

  • operating outside the 5, 9 and 14 Eyes intelligence alliances
  • an independent no-logs audit, worth more the newer it is and a third of the points once it is over three years old
  • RAM-only servers
  • a written no-logs policy
  • accepting anonymous payment

Medium and high severity incidents from the last five years take points away. Low severity events appear on the review without a penalty, for example a server seizure that gave up no customer data.

Security features

Security points go to:

  • WireGuard and OpenVPN support
  • a kill switch on both desktop and mobile
  • obfuscation for networks that block VPNs
  • multi-hop connections
  • post-quantum encryption
  • the provider’s own DNS servers with leak protection
  • proper IPv6 handling
  • an independent audit of the apps or servers in the last three years

Price and value

The price score looks at six things:

  • the lowest effective monthly price across plans
  • the month-to-month price
  • the length of the money-back guarantee
  • a free tier
  • the renewal price
  • the choice of payment methods

A plan with no published monthly price earns nothing for the month-to-month price. Renewal is scored from what the official page says:

  • A stated renewal amount is compared with the first-term total.
  • A plan that renews at the same price earns the full renewal point.
  • A renewal the page does not state earns half because we never guess a renewal price.

The other five dimensions

  • Speed and reliability will use our lab’s figures for throughput, latency, connection time and uptime. The lab is not running yet, so this dimension is left out for now.
  • Streaming and unblocking treats a service the provider documents as a claim until we test it. A tested success counts for more. A Smart DNS service for devices that cannot run a VPN app adds a point.
  • Apps and usability covers the platforms, the number of devices at once, router support, browser extensions, open-source apps and split tunneling.
  • Support covers live chat, email and a knowledge base. Live chat around the clock scores highest, and we will add measured response times once we have them.
  • Transparency covers transparency reports, a warrant canary, open-source apps and the number of independent audits.

Missing data and score changes

A dimension with no data does not count as zero. We leave it out and scale up the other weights so the total still runs from 0 to 10.

Each score shows how much of the weight it covers. Today that is 80 percent for every provider because speed is missing. When the lab has 30 days of data for a provider, its speed dimension switches on.

Scores are recalculated from our provider data every time the site builds. A price change, a new audit, a documented incident or a new feature moves the score at the next build. Our open data page lists the latest data changes. Nobody edits a score by hand, and the build fails if the published scores do not match a fresh calculation.

Changes to the rules

The weights and rules change only after a review by the editorial team. Each change gets a dated entry in this page’s changelog that says what changed and why. Affiliate deals are never part of the score. A provider with no affiliate program is scored exactly like one with a program.

Editorial and AI policy

Our principles

  1. Reader first. A page exists to give the best answer to its question, as a skeptical expert who has used the products would judge it.
  2. Data over adjectives. Every superlative comes with a number, a date and a source.
  3. The bad parts too. Every review and best-of page lists real drawbacks and says who should not buy.
  4. Nothing invented. We never make up test results, quotes, user reviews, screenshots or dates. If we have not tested something, the page says so.
  5. Commission never changes a verdict. Rankings come from the scoring rules on this page.
  6. Dates are earned. The “updated” date changes only when the substance of a page changes.

Sources

Facts about a provider come from:

  • its official pages
  • audit reports
  • regulators, laws and court documents
  • our own research
  • major news outlets with named reporters

We never use other review sites as a source for facts. Every page lists its sources with the date we read each one. The provider record behind each page stores the official link for every section.

How AI helps the team

An AI system called Claude does the research, monitoring, drafting and routine updates on this site. It works under rules set by the VPN Clue Editorial Team, and automated checks enforce those rules. A page cannot be published when it:

  • lacks required sections, sources, disclosure or depth
  • uses banned marketing phrases
  • links to pages that do not exist
  • changes a date without a logged reason

The AI reads official pages and records what they say with the link and the date. From that data it drafts and updates our pages. It also watches for new prices, plans, audits and laws and runs the technical checks.

The AI does not invent facts, scores, quotes or test results. It cannot change the scoring rules, and it cannot publish a page that fails the checks.

The editorial team is responsible for everything we publish. Each page written with AI help says so in the box at its foot.

Sensitive topics

We cover laws, bans and age-verification rules factually and neutrally, and nothing on this site is legal advice. Our pages carry no instructions aimed at minors and no links to adult sites. We never encourage anyone to break the law or a service’s terms, and our torrenting and media-player guides carry a copyright notice.

We do not write about using a VPN in countries that ban or restrict VPNs or punish people for using one. A mistake there could put a reader at real risk, and the rules can change without warning. Our country pages cover only places where using a VPN is legal. If you plan to use a VPN somewhere else, check your own government’s official travel advice first.

How we make money

Some links on VPN Clue are affiliate links. If you click one and buy a subscription, the provider pays us a commission. You pay the same price as you would by going to the provider directly, and the money comes from the provider’s marketing budget.

Every affiliate link on the site:

  • goes through an address that starts with /go/
  • carries the sponsored attribute that search engines require
  • sits below a visible disclosure on the same page

We have affiliate deals with NordVPN, Surfshark, Proton VPN and ExpressVPN. Where we have no deal, the link goes straight to the provider’s website. Mullvad and IVPN run no affiliate programs at all. We review, score and recommend them on exactly the same terms as everyone else.

How we keep money out of the rankings

  • Scores come from the published rules on this page. Whether a provider pays us is not an input, and nobody can edit a score by hand.
  • Best-of lists follow the criteria for that use. Each pick shows the data that put it there.
  • Every review lists drawbacks and who should skip the service, even for the providers that pay the highest commissions.
  • We read prices and deals on official pages and show the date we checked them. We do not list expired coupons or invented discounts.
  • Affiliate programs never see or approve a page before it goes live, and we do not run sponsored posts.

What we do not do

We do not sell rankings or placements, and we do not run display ads yet. If we ever add ads, they will be labeled and placed below the main content.

We do not collect or sell reader data. Our tools store nothing you type and nothing they measure. The privacy policy explains what our hosting provider handles when it serves a page.

The disclosure sits in plain words above the first affiliate link on each page instead of in the footer. Regulators in the United States and the United Kingdom require clear disclosure of paid links, and we think readers deserve it anyway.

Corrections

How to report an error

Send these three things to corrections@vpnclue.com:

  • the page address
  • the sentence or number that is wrong
  • an official source that shows the right value

The best source is the provider’s own page, an audit report, a law or a court document. A screenshot with a visible date also helps. Our contact page lists our other addresses.

Providers can also report errors, and we treat their reports like any reader’s. A claim needs a public source we can cite. We do not change a fact on the strength of an email alone.

What happens next

  1. We check the claim against the source within 48 hours.
  2. If the page is wrong, we correct it and add a dated changelog entry that says what changed. When the error came from our data, we also fix the provider record.
  3. If the claim does not hold up, we reply with the source we relied on.
  4. If a correction changes a score or a ranking, the affected pages update at the next build.

What counts as an error

An error is any fact we got wrong, from a price or a server count to an audit detail or the name of an owner. We welcome disagreement with a verdict as well. We treat it as feedback unless it rests on a fact we got wrong. Typos and broken links get fixed as soon as we see them, without a changelog entry.

Each correction appears in the changelog at the foot of the page it affects, with the date and a short note. Many pages have carried such entries since the site launched on 31 October 2026. A site that tracks hundreds of changing prices and policies will make mistakes. What we promise is to fix them quickly and visibly and never to edit a page quietly to hide one.

Changelog

  • Merged five pages into this one: How we test, Scoring methodology, Editorial and AI policy, How we make money and Corrections. Added our rule on countries where using a VPN puts people at risk.
  • Corrections: added the address for reports, corrections@vpnclue.com. Corrected the corrections log, because many pages have recorded corrections in their own changelogs since launch.
  • Scoring: corrected two descriptions to match the scoring code. A no-logs audit older than three years still earns a third of the audit points. Recent data changes are listed on the open data page instead of each page's changelog.
  • Testing: corrected four statements to match how the site works today. Tables and charts come from the same data, and a daily check flags text that no longer matches. Pages show one date, data changes are listed on the open data page and the lab is still planned.
  • Published our testing method, scoring method version 1 with speed left out until the lab runs, the editorial and AI policy, how we make money and the corrections process.
  • Scoring: renewal is now scored from what the official page states. A stated amount is compared with the first term, the same price earns the full point and nothing stated earns half. A plan with no published monthly price earns no monthly points.