A recent study challenges the common assumption that all VPNs provide a uniform view of the Internet when measuring web traffic from a specific economy. Researchers found that changing VPN providers can drastically alter which servers a browser connects to and where those servers are geolocated, leading to conflicting conclusions about domestic infrastructure reliance and cross-border data flows.
To investigate this phenomenon, the team developed a system called VPN-Scope, which simultaneously visited the same websites through different providers using connections from the same economy and city. The study analyzed four major providers—ExpressVPN, NordVPN, ProtonVPN, and Surfshark—across 14 economies in Europe and the Americas. Over a three-week period in March and April 2025, the system performed 25 visits to 1,000 selected websites per economy, recording server locations for both main pages and embedded resources like scripts and fonts.
The results highlighted significant discrepancies depending on the provider chosen. In Chile, for instance, only 9.5% of requests appeared to be served domestically when using ExpressVPN. In contrast, Surfshark and NordVPN showed domestic service shares of 62.7% and 64.2%, respectively. Furthermore, the distribution of traffic to other economies varied wildly; Brazil accounted for 57.1% of traffic through ExpressVPN but only 2.4% through NordVPN. These differences suggest that a study using one provider might conclude an economy is largely self-sufficient in hosting, while another provider could depict the same economy as heavily dependent on foreign infrastructure.
DNS experiments further revealed that approximately 12% of initial domain lookups and 31% of dependency domains returned different address sets when comparing two providers in the same economy. However, these address differences did not always translate to different hosting organizations. In 98% of initial-domain comparisons and 96% of dependency-domain comparisons, the DNS answers shared at least one hosting organization. This indicates that while VPN choice strongly affects the perceived geographic location of content, it has a more stable impact on identifying which companies serve that content.
The study attributes these variations to factors beyond simple browser settings, including Content Delivery Network (CDN) steering and network routing decisions. Even when connections originated from the same city, such as Milan, they could reach physical servers in different locations like Frankfurt due to how networks route traffic. Consequently, researchers are advised to treat the VPN provider as a critical component of experimental design rather than a neutral tool.
Experts recommend that future studies report specific provider and connection settings, verify actual exit node locations, and consider repeating measurements across multiple providers when geographic conclusions are paramount. The findings underscore that results obtained through a single VPN should be viewed as one perspective of the network rather than an objective, provider-independent picture of an economy's digital infrastructure.
Sources: APNIC Blog.