How we review — and why you can trust the order

Every other "best VPN" list ranks whoever pays the biggest commission. We don't. Here's exactly what we measure, how we score it, where the data comes from, how we make money, and how often we refresh — in plain language, with nothing hidden.

Why this site exists

The author worked inside a major VPN company and saw the mechanic firsthand: mainstream rankings track commission size, not product quality — the highest bidder wins the #1 slot. VPN Insider inverts that. The ranking input is real user sentiment at scale, and commission is structurally barred from influencing position. If a product is good, the people who use it say so; our job is to measure that honestly and show our work.

What we measure — and why it matters

How we score sentiment

Each VPN's headline number is a blend of two independent sources:

1. Power-user sentiment (the discriminating signal)

We gather public discussion from communities where people know VPNs (Hacker News today; more sources being added), then an LLM reads sentiment toward the named VPN, theme by theme. We use an LLM rather than keyword scoring for a concrete reason: in our own testing, a simple lexicon scorer agreed with the true meaning only ~40% of the time — it reads "I would never use X" or sarcasm as positive. The LLM doesn't.

2. Mainstream rating (the volume signal)

Apple App Store lifetime star ratings, pulled via Apple's official API. Huge sample sizes — but compressed (almost every VPN sits near 4.5★) and prompt-inflated, so they confirm popularity, not quality. That's why we weight them less, and why the gap between the two sources is itself a finding: a big gap means casual stars over-flatter a VPN the knowledgeable distrust.

What we filter and how we weight

• Spam & scam posts are dropped; promotional / astroturf posts are down-weighted, not deleted.
• Recency matters — newer posts count more (12-month half-life).
• Usernames are never published; we quote sparingly, with a link to the source.
• Thin data is labeled, not faked. Below a minimum volume a VPN is marked limited data or its score is withheld (see PIA and IPVanish).

Category scores (speed, streaming, etc.) blend our sentiment data with independently reported testing from labs such as Security.org, CyberInsider, and vpnMentor. Where a number comes from testing rather than sentiment, the review says so. We don't yet run our own speed lab — when we do, we'll disclose the rig.

Where the data comes from

Everything is gathered through official APIs or licensed feeds — never bulk-scraped. Scraping would violate platform terms and, more importantly, undermine the credibility this whole site is built on. Prices are verified against vendors' live pricing pages and dated. Our reproducible scoring pipeline and the lexicon-vs-LLM test are documented in our research notes.

How we make money — and the firewall

The VPNs at a glance

Click any column heading to sort. "True 3-yr cost" includes the renewal hike. Sentiment "—" means score withheld for limited data.

What we're still improving (honestly)

• More sources. We're adding app-store review text, and Reddit/Trustpilot once approvals and licensing clear — to widen beyond the technical Hacker News skew.
• Our own speed lab. Today's speed/streaming figures come from cited third-party testing; we'll disclose our own methodology when we run it.
• Scoring two of nine. PIA and IPVanish don't yet have enough clean signal to rank — we'd rather withhold than fake precision.
• Reader submissions. A moderated review form is planned; submissions will be a clearly-labeled separate source until abuse-handling is proven.

VPNInsider.org · How we review · updated monthly · this cycle June 2026.