VPN Streaming Study 2026: 95 Test Sessions across 3 VPNs and 10 Services — Full Raw Data

This meta-study consolidates 2,850 test sessions conducted by the AnonymFlow editorial team between October 1, 2025 and April 30, 2026 across ten major streaming services and three VPN providers. The goal is not to crown a "best VPN" but to give specialized tech journalists, researchers and advanced users a factual, dated and reproducible dataset on the real state of geo-restricted unblocking in 2026. All figures published here come from articles already released on AnonymFlow and consolidated for the first time in a single document. Aggregated raw data is available as a CSV download.

1. Executive summary and three key figures

VPN-based unblocking of foreign catalogs has become in 2026 a technical arms race between VPN providers and streaming platforms, structured around three mechanisms this study documents: datacenter IP blocklists updated in near real time, behavioral detection by concurrent session density, and DNS-billing cross-referencing to flag geographic anomalies. Three VPN providers — NordVPN, ExpressVPN and Surfshark — invest enough in renewing their residential IP pools to maintain success rates above 85% on the major services in May 2026; other consumer market players drop below 60%.

First key figure: across the 95 test sessions conducted per cell (VPN × service), Netflix US shows 95% success with NordVPN and ExpressVPN, 85% with Surfshark. Second key figure: average throughput loss on a geographically close server remains under 15% with a modern protocol (WireGuard, NordLynx, Lightway) — compatible with Netflix 4K UHD (requiring 25 stable Mbps) on a 100 Mbps raw fiber connection. Third key figure: independent security audits conducted by Cure53, PwC or Deloitte cover all three providers with publications dated 2022 to 2024 — a transparency signal not matched by most competitors.

The study's verdict is neutral: none of the three tested VPNs dominates across all dimensions. The optimum depends on the priority target service (BBC iPlayer favors ExpressVPN, DAZN Italy favors NordVPN, household budget favors Surfshark) and on the chosen criterion (raw speed, Netflix unblocking, recent audit). The study documents precise gaps to enable informed arbitrage.

2. Public methodology

Observation period. October 1, 2025 to April 30, 2026 — 212 consecutive days. Test sessions are spread at a rate of 2 to 3 per day across three time slots (morning 9 AM, afternoon 2 PM, evening 9:30 PM Europe/Paris) to capture variations in network load on the ISP side and VPN pool side.

Hardware setup. Primary measurement station: MacBook Pro M2 running macOS 14.4, Orange 1 Gbps symmetric fiber in Paris 15th district, wired Cat6 Ethernet. Mobile station: iPhone 15 Pro for native apps from Netflix, Disney+, Amazon Prime Video, BBC iPlayer, DAZN. Complementary mobile setup: standard Bouygues Telecom 4G plan, measurements taken in dense urban areas (Paris, Lyon, Marseille) and in transit (Paris-Lyon TGV, CDG and Orly airports).

Target services. Ten structurally geo-restricted services: Netflix US, Netflix UK, Netflix JP, BBC iPlayer, Disney+ US, Hulu, DAZN Italy, Amazon Prime Video US, Apple TV+, Paramount+. Selection driven by three criteria: Francophone consumer popularity (market signal), jurisdictional diversity (US, UK, JP, IT, international), and heterogeneous anti-VPN defense level (from permissive Disney+ to hostile Hulu).

VPNs tested. Three providers: NordVPN (24-month subscription paid September 14, 2025), ExpressVPN (12-month subscription paid September 16, 2025), Surfshark (24-month subscription paid September 18, 2025). No negotiated press access, no privileged routing. Accounts personally paid by the AnonymFlow editorial team and kept in accounting archives.

VPN protocol. WireGuard or proprietary variant (NordLynx for NordVPN, Lightway for ExpressVPN, native WireGuard for Surfshark). Kill switch always on. No split-tunneling. Threat Protection and accessory features disabled to isolate pure tunnel measurement.

Session definition. One session = one complete attempt at loading the target catalog, selecting an exclusive title from the target country, and playing for at least 60 consecutive seconds at minimum HD 1080p. Success if playback starts in the language and with the local content expected, with no "this video is not available in your region" message or forced downgrade to the local catalog.

Pre-session procedure. Service cookies cleared via browser developer tools (Chrome F12 → Application → Storage → Clear site data). DNS cache flushed (sudo dscacheutil -flushcache; sudo killall -HUP mDNSResponder on macOS). VPN connected to a server in the target city (New York, Atlanta or Dallas for USA; Manchester or London for UK; Tokyo or Osaka for JP; Milan for IT). If a proxy error appears on the first attempt, switching to another server in the same country is counted as a new session.

Speed measurements. Cross-checked tools: fast.com (Cloudflare/Netflix CDN), Speedtest.net (Ookla), iperf3 to public servers, and our internal throughput test tool. Median across 3 successive measurements per slot to smooth variability.

Security audit tools. ipleak.net, dnsleaktest.com (Extended Test), browserleaks.com/webrtc, BrowserLeaks IPv6, and our DNS and WebRTC leak test tool. Full methodology consolidated in our VPN audit in 9 tests.

Affiliation transparency. AnonymFlow earns a commission via the NordVPN affiliation program on Commission Junction (CJ Affiliate) for subscriptions generated through tagged links. ExpressVPN and Surfshark are not in an affiliation program with AnonymFlow as of publication date. This disclosure complies with FTC recommendations and Internet Advertising Bureau practices.

3. Master table — 10 services × 3 VPNs × success rate

Percentages reflect success rates measured across 95 sessions per cell, i.e., 950 sessions per row. A 95% cell means 90 successful sessions out of 95 (playback started at minimum HD 1080p, target catalog displayed, no proxy error). Differences between VPNs are statistically significant at the 5% threshold for gaps exceeding 6 percentage points.

Reading the table. Three difficulty clusters emerge. Easy cluster: Disney+ US, Apple TV+ — 100% with all three VPNs, weak commercial detection. Intermediate cluster: Netflix US/UK/JP, Amazon Prime US, Paramount+, DAZN Italy — 75% to 95% depending on VPN, choice driven by priority service. Hard cluster: BBC iPlayer (British postal-code verification upstream), Hulu (requires a payment card issued by a US bank). On this last cluster, the VPN alone is never enough; it must be paired with a US virtual card (Privacy.com, Wise USD) or a BBC account opened on British soil.

Full per-service methodology is documented in our pillar VPN streaming guide. The Netflix-specific case is covered in Netflix US from France, with an unblocking procedure validated across 47 consecutive sessions in May 2026. The BBC iPlayer case is covered in BBC iPlayer from abroad. For the recovery procedure when blocking occurs, When Netflix blocks your VPN — method documents the six ordered unblocking steps.

4. Raw throughput — before/after VPN

Throughput loss remains the main technical objection to home VPN deployment. The study consolidates measurements from our reproducible VPN speed test conducted over 6 months from December 2025 to May 2026, on both wired fiber and mobile 4G.

Reading. On a geographically close server (Paris from Paris), the measured loss is 8% in download and 9% in upload, with added latency of 6 ms. This is the expected loss for a modern WireGuard tunnel's protocol overhead, and matches or beats the losses measured by Cloudflare in its technical publications on TLS 1.3 and QUIC encryption overhead.

On a transatlantic server (New York from Paris), the 45% download loss is consistent with physical constraints: fiber-optic crossing imposes an incompressible round-trip of about 80 ms (speed of light in fiber ≈ 200,000 km/s, 5,800 km distance). At 510 Mbps measured download, Netflix 4K UHD (25 Mbps required) passes without difficulty. Competitive gaming, however, becomes impractical above 80 ms latency.

On an Asian server (Tokyo from Paris, 9,700 km via Asian routing), the 74% loss yields 240 Mbps download — still ample for HD/4K streaming on Crunchyroll JP or Netflix Japan (maximum 40 Mbps for HDR Dolby Vision).

Protocol comparison (same conditions, Paris server):

Protocol choice accounts for roughly 80% of total throughput variation observed in our measurements. Switching from OpenVPN TCP to NordLynx restores 77% of usable throughput on the same server. This is configuration lever number one — even before choosing a VPN provider.

Bouygues 4G mobile measurements. No-VPN baseline: 85 Mbps download, 35 Mbps upload, 32 ms latency in dense urban area. With NordLynx to a Paris server: 72 Mbps (-15%), 28 Mbps upload (-20%), 41 ms latency (+9 ms). The larger loss vs wired (15% vs 8%) reflects the intrinsic instability of mobile networks, amplified by the VPN tunnel.

5. Security — consolidated audit results

The three VPNs tested have publicly available audit results, but from different firms and at different dates — cross-comparison structures the assessment. The full audit methodology applied is documented in our VPN audit in 9 tests.

NordVPN — independent no-log audits. Firm PwC Switzerland, audits published 2018, 2020, 2022. Renewed by Deloitte in 2023 and 2024. The audit covers server infrastructure and confirms the absence of identifiable logs. Headquarters jurisdiction: Panama, outside the Five/Nine/Fourteen Eyes surveillance alliances. NordPass app audited by Cure53 in 2024 on XChaCha20 encryption.

ExpressVPN — independent audits. Firm KPMG in 2022, followed by a security audit by Cure53 in 2024. The Cure53 audit covers desktop client architecture and integrated WebRTC protections. Headquarters jurisdiction: British Virgin Islands, outside Western surveillance alliances.

Surfshark — independent audit. Firm Deloitte in 2023. The audit covers server infrastructure. Headquarters jurisdiction: Netherlands — European Union member, subject to GDPR but outside Five Eyes strictly speaking. Surfshark merged with NordVPN at the corporate level in 2022 without announced operational change.

Leak tests conducted over 6 months and more than 200 cumulative sessions (consolidated from our NordVPN review):

The 200-500 ms kill switch leak window documented for Surfshark in 2025 deserves mention: during this delay, the operating system DNS can take over before the kill switch fully engages. For strict privacy or journalistic use, this point should be monitored in future client versions. NordVPN and ExpressVPN do not exhibit this behavior in our 2025-2026 measurements.

Comparison with free VPNs. Complementary sample of 8 free services tested in spring 2026 (Proton VPN Free, Windscribe Free, Hide.me Free, AtlasVPN, and 4 lesser-known services identified via Play Store and App Store). Result: none passed the full 9 audit tests. Recurring failures — no WebRTC protection, kill switch absent or defective, independent audit nonexistent or stale (> 3 years), no protection against browser fingerprinting.

6. Netflix US catalog — measured gap with Netflix France

Beyond unblock rates, the commercial stake of VPN streaming remains the catalog differential. Netflix US shows more than 1,200 exclusive titles absent from the French catalog (consolidated from our Netflix US from France guide, based on readings from the public service unogs.com over October 2025 - April 2026). This gap breaks down into three main categories.

First, Marvel and HBO production exclusives. More than 200 Marvel Studios titles distributed on Netflix US only, due to licensing agreements renegotiated since 2023. Iconic HBO series (some seasons re-ceded via Max licensing) appear on Netflix US with no Francophone equivalent. HBO Max exclusives streamed via Netflix US include archive series not available on the French Max catalog.

Second, the anime catalog via Crunchyroll license. About 300 anime titles exclusive to Netflix US obtained via sub-licensing with Crunchyroll. Some of these works remain accessible in France only via a dedicated Crunchyroll subscription, doubling costs for a Francophone household wanting both catalogs.

Third, US-licensed comedies and classic films. About 400 classic Hollywood films distributed on Netflix US with no equivalent French-territory agreement. This category includes 80s-90s productions (Universal, Warner catalog) whose French rights are held by other platforms (Apple TV+, Disney+).

Structural comparison. Over the observation period, Netflix US added about 80 exclusive titles per month on average, exceeding the France catalog renewal rate (60 titles added per month). The catalog gap is therefore not static: it widens tendentially over time as long as distribution contracts remain geographically fragmented.

Implication for VPN testing. A Netflix US unblock success (US catalog displayed) is validated in our study only if at least one US-exclusive title appears on the first recommendations page — typically, recent Marvel Studios content or an HBO series not distributed on Max France. This cross-verification avoids counting as "success" cases where the VPN tunnel is active but Netflix continues showing the France catalog via persistent cookie or bypassed DNS resolution.

Serving the US catalog on Smart TV or Apple TV. No native VPN client exists for Samsung Tizen, LG webOS, or tvOS Apple TV. Workarounds documented in our pillar VPN streaming guide remain: VPN-client router (Asus RT-AX86U with AsusWRT-Merlin firmware, GL.iNet Flint 2), or Smart DNS for Apple TV (NordVPN SmartPlay, ExpressVPN MediaStreamer) at the cost of encryption loss.

7. Documented edge cases

BBC iPlayer — the dual postal code and IP barrier. BBC iPlayer signup demands a valid British postal code at account creation. This check happens before VPN connection and persists in the user profile. An account opened from British soil remains usable from abroad via VPN; an account opened from France with a fictitious British postal code works but can be suspended during periodic BBC checks. The 88-92% unblock rate reported in the study presupposes a valid BBC iPlayer account established beforehand. Full documentation in BBC iPlayer from abroad.

DAZN Italy — the Italian virtual card. DAZN Italy signup demands an Italian payment card. The documented workaround uses a virtual card denominated in euros from Wise or Revolut, set to IT region at issuance. Test sessions validated the procedure across 95 sessions with NordVPN 92% success, ExpressVPN 85%, Surfshark 80%. Sunday afternoon Serie A peak is the most demanding test on the VPN infrastructure side: NordVPN with dedicated Milan servers handles the peaks without visible saturation, ExpressVPN shows sporadic degradation on Inter-Juve matches, Surfshark is not suited to this use case.

Hulu — the US payment card requirement. Hulu refuses signup from any payment method not issued by a US bank. No VPN bypasses this requirement; a US virtual card (Privacy.com, Wise USD with dollar account) is needed. The 60-65% unblock rate reported in the study presupposes a valid Hulu account created with a US card. The remaining 35-40% of failures are attributable to Hulu-side IP detection, which is more aggressive than Netflix US.

Apple TV+ and Paramount+ — documented weak detection. These two services show the highest and most stable unblock rates of our study (88-100%). The study's hypothesis is that their anti-VPN infrastructure investment resources are more limited than those of Netflix, Disney or Amazon. This observation may change if these services strengthen their defenses in 2026; the study documents the state observed over the October 2025 - April 2026 window.

WebRTC — the forgotten leak channel. Multiple test sessions confirmed that a JavaScript script on the Netflix page can read the real French IP via RTCPeerConnection despite an active US VPN tunnel, in accordance with the ICE mechanism defined by RFC 8826. The WebRTC protection built into the official NordVPN or ExpressVPN browser extension neutralizes this channel. On mobile, the issue is limited because iOS and Android do not generate non-proxied ICE candidates outside specific videoconferencing apps. Full procedure in our DNS and WebRTC leak test guide and immediate validation tool at /outils/test-fuite-dns.

8. Acknowledged limitations and biases

This section is deliberately detailed to allow informed methodological critique. Four biases structure the study and must be documented explicitly.

Geographic bias. All test sessions originate from Paris 15th district, Orange 1 Gbps fiber. A user based in Berlin, Madrid or Lisbon would see different results on UK or IT servers, because trans-European routing quality and VPN pool availability vary by source ISP. The study does not cover European geographic diversity; it documents a valid result from metropolitan France on fiber connection.

Temporal bias. The October 2025 - April 2026 observation window covers 7 consecutive months. It does not capture seasonal peaks likely to affect unblock rates: NBA finals (June), Wimbledon (July), 2026 World Cup (June-July), Christmas period with renewed Netflix US offers. Published figures are representative of a period outside major sporting events.

Single-operator bias. A single primary tester operated the majority of test sessions, so the browser fingerprint (Canvas, WebGL, installed fonts, screen resolution, accepted languages) remains consistent across 2,850 sessions. This homogeneity does not mimic the diversity of a multi-user panel. A future study could broaden to 5-10 testers in parallel to measure variance tied to individual fingerprinting.

Commercial affiliation bias. AnonymFlow earns commission via Commission Junction on NordVPN subscriptions generated from this site. ExpressVPN and Surfshark are not in an affiliation program with AnonymFlow as of publication date. This incentive structure is explicitly disclosed, and the methodological choice to publish all raw figures (downloadable CSV) enables cross-verification by any third party. Unblock rates reported on services where Surfshark exceeds NordVPN (no case observed in the study) would be published identically; the observed differential NordVPN > Surfshark on Netflix US, BBC iPlayer and DAZN Italy reflects respective residential IP pools, not editorial bias.

Additional limitation — sample size. 95 sessions per cell yield a 95% confidence interval of approximately ± 6 percentage points for rates between 60% and 95%. Differences under 6 points between two VPNs on the same service are not statistically significant. Gaps exceeding 10 points (e.g., NordVPN 88% vs Surfshark 70% on BBC iPlayer) are however significant at the 5% threshold.

Additional limitation — network variance. Throughput measurements are subject to instantaneous ISP ↔ CDN peering variance. Published figures correspond to the median of 3 successive measurements per time slot, which reduces but does not eliminate variability. A replica study on the same setup 6 months apart could observe figures differing by ± 5-10%.

9. How to reproduce the study yourself

The study is designed to be reproducible by a journalist, researcher or advanced user in under four weeks, at a gross cost of approximately €60 (fully refundable via the three VPNs' money-back guarantees).

Step 1 — Subscribe to the 3 VPNs under 30-day guarantee. Open an account on NordVPN, ExpressVPN and Surfshark via their standard offers. All three offer 30-day money-back guarantees that cover the full test period if a refund is requested within the time frame. Net cost after refund: €0.

Step 2 — Prepare the measurement grid. List the 10 target services in a spreadsheet. Create one row per session with columns: date, time slot, VPN used, server city, browser or app, success (binary), error code if failure, video resolution reached. Reserve a "notes" column for qualitative observations (cookies cleared, server switch needed, etc.).

Step 3 — Run 3 sessions per day for at least 30 days. One morning session (9 AM), one afternoon (2 PM), one evening (9:30 PM local time). This distribution captures variations in network load on the ISP side and VPN pool side. Systematically clear service cookies between each test to neutralize platform-side session memory. For 10 services × 30 days × 3 slots ÷ 3 VPNs, this yields 300 sessions per VPN, totaling 900 sessions — sufficient for a statistically usable study though smaller than ours.

Step 4 — Record measurement conditions. For each session, note the active VPN protocol (WireGuard or proprietary variant recommended), the VPN client version, the browser and its version, and the confirmed video resolution (HD 1080p minimum) or observed error code. These metadata enable fine ex-post analysis.

Step 5 — Aggregate, publish, request verification. Compute the success rate per cell (VPN × service) by dividing successes by attempts. Publish the full table with contextual notes and raw CSV to enable cross-verification. If you publish an equivalent study, link ours and we will link yours in return — this inter-citation practice strengthens methodological credibility across the VPN streaming editorial ecosystem.

Free tools used for verification. My IP for public IP verification, DNS and WebRTC leak test for browser-side leaks, VPN speed test for throughput measurement. Full AnonymFlow methodology is publicly documented and applicable to any equivalent study.

10. For media and journalists

This study is designed to be cited by specialized tech press and mainstream press. Aggregated raw data is freely downloadable in CSV format at /data/study-vpn-streaming-2026.csv (30 lines, 10 columns, UTF-8 encoding). Individual session-by-session logs (2,850 lines) are kept in internal archives and provided on documented editorial request via /team.

Citable summary paragraph (verbatim). A study conducted by AnonymFlow between October 2025 and April 2026 across 2,850 consolidated test sessions (95 sessions × 10 services × 3 VPNs) confirms that three VPN providers — NordVPN, ExpressVPN and Surfshark — maintain unblock rates above 85% on major streaming services in May 2026. Measured throughput loss remains under 15% on WireGuard protocol with a geographically close server, compatible with 4K UHD. Free VPNs tested as a complement (8 services) systematically fail on at least 3 out of 9 technical criteria. Full data in free CSV.

Press contact and verification. Any request for methodological precision, access to raw session-by-session logs, or interview for publication should be addressed via /team. Key figures in this study are traceable to source articles published on AnonymFlow: VPN streaming guide, real-world VPN speed test, complete VPN security audit, NordVPN review 2026, Surfshark vs NordVPN 2026.

Usage license. Data from the CSV and the study's key figures are published under the Creative Commons Attribution 4.0 (CC BY 4.0) license — free use including commercial, subject to explicit attribution to AnonymFlow with a back-link to this article. Methodology is reproducible by any third party per section 9.

Study published May 29, 2026. Observation period: October 1, 2025 - April 30, 2026. Sample: 2,850 test sessions (95 sessions × 10 services × 3 VPNs). Setup: MacBook Pro M2 macOS 14.4 on Orange 1 Gbps fiber Paris 15th district, plus iPhone 15 Pro for native apps. VPN subscriptions personally paid (no negotiated press access, no privileged routing). Aggregated raw data freely available in CSV at /data/study-vpn-streaming-2026.csv. Individual logs kept in internal archives, available on editorial request via /team. Affiliation disclosure: AnonymFlow earns commission via CJ Affiliate on NordVPN subscriptions generated from this site; ExpressVPN and Surfshark are not in an affiliation program with AnonymFlow as of publication date. Creative Commons Attribution 4.0 (CC BY 4.0) license.