Criteria that actually matter
Jurisdiction & ownership
Where the company is legally based determines which governments can compel data. Prefer providers outside 14-Eyes intelligence-sharing jurisdictions, with publicly named ownership — anonymous "offshore" providers are a red flag, not a feature.
No-logs — proven, not promised
"No logs" is a claim; look for evidence: independent audits (PrivacyTools lists audited providers), court cases where providers had nothing to hand over, or warrant canaries. Proton VPN's explainer covers what genuine no-logs architecture looks like.
Protocol & implementation
WireGuard for speed, OpenVPN for compatibility; both fine if correctly implemented. Avoid providers pushing proprietary protocols with no third-party audit.
Kill switch & leak protection
A kill switch that actually works (verify it — disconnect mid-test), plus DNS/IPv6 leak prevention. Test on ipleak.net before trusting it.
Criteria that mostly don't
- Server count — 3,000 virtual servers on the same hardware means little; latency to your routes matters more.
- "Military-grade encryption" — AES-256 is a marketing term; every serious VPN uses it.
- Obfuscated server branding — useful in censored regions, irrelevant elsewhere.
- Anonymous payment options — nice bonus, but the provider's jurisdiction matters far more than how you paid.
Realistic expectations
- Speed loss: 10–30% typical on good routes; more on distant servers.
- Not anonymity: a VPN shifts trust from your ISP to the provider — it doesn't make you invisible.
- Not legality: what's illegal without a VPN stays illegal with one.
- Streaming/torrenting: varies by provider and changes monthly; test before committing to a long plan.
How to test before paying
- Use the trial/money-back window — measure speed at your real usage times.
- Run ipleak.net and dnsleaktest.com — confirm no DNS or IPv6 leaks.
- Test the kill switch under load — disconnect mid-download, check for leaks.
- Read recent independent reviews — check for ownership changes and fresh audit results, not just old rankings.
Sources & further reading
- PrivacyTools — VPN criteria and recommendations
- Proton VPN — what no-logs actually means
- TorrentFreak — annual VPN survey
VPNs — thirty years from corporate plumbing to consumer promise
The VPN began as corporate infrastructure: PPTP (1996) let Microsoft's clients tunnel into office networks; IPsec and OpenVPN (2001) made encrypted site-to-site the standard. The consumer product emerged in the late 2000s when public Wi-Fi made "encrypt my coffee-shop traffic" a mass need, and the marketing machine built the modern "trust us with everything" industry.
The trust era broke twice. First, providers got caught logging — HideMyAss handing user data to the FBI (2011) and several "no-logs" services turning out to log. Second, the audit era (2018+) let third parties actually verify claims — Deloitte, PwC and Cure53 audits became the industry's answer to "why should we believe you".
What's actually verifiable, counted
| Feature | Reality check |
|---|---|
| WireGuard vs OpenVPN code size | ~4k lines vs ~100k+ — auditability is a feature, not a buzzword |
| WireGuard speed advantage | roughly 2-4× OpenVPN throughput in most real tests |
| Top providers with published audits | majority of the top-10 have at least one public audit by 2025 |
| "No-logs" claims verified vs. marketing | a minority hold up under audit or court discovery — the rest are vibes |
| Free VPN data monetization | documented selling of user bandwidth/data in multiple free tiers |
The filter that works: a VPN shifts trust to its operator, so pick the one with the most to lose — audited no-logs, legal entity you can sue, ownership that's public. The marketing metric that doesn't predict anything is "military-grade encryption": they all use the same AES-256.
The questions buyers actually ask in 2026
From search data and forum threads, the modern VPN buyer's query list has narrowed: does it keep logs (the audit question), does it work on my devices (the protocol question), does it slow me down (the WireGuard answer: rarely), and does it unblock what I want (the cat-and-mouse question no provider can guarantee). The marketing questions — "is it the fastest", "is it military-grade" — have largely faded as buyers got literate.
Testing a provider yourself
The home audit is simpler than review sites suggest: baseline your speed without VPN, re-run through 3-4 servers at different times, check for DNS/WebRTC leaks on any of the standard test sites, and read the actual privacy policy for data retention wording — "no logs" marketing means nothing next to the legal text. An honest provider's policy is boring and specific; a dishonest one's is vague about what it collects "to improve service".
The market's center of gravity has shifted to audited no-logs + open-source clients + WireGuard-class performance. Any provider still selling on 2015-era claims ("military encryption!") is telling you where their budget goes.
Expert context · 2026
“A VPN shifts your trust from your ISP to the VPN provider — it doesn't remove it. The question is always who you trust more.”
— Bruce Schneier, security technologist and author · schneier.com