Research from four US and Canadian universities shows that, from April to Dec. 2020, China's Great Firewall blocked ~311K domains, ~41K seemingly by accident
Catalin Cimpanu / The Record : Tweets: @therecord_media Tweets: @therecord_media : In the largest study of its kind, a team of academics from four US and Canadian universities said they were able to determine the size of China's Great Firewall internet censorship capabilities https://therecord.media/...
Context & Ripple Effects
This study is the baseline measurement in an arc of Great Firewall research: a team from four US and Canadian universities quantified the system's full domain-blocklist scale for the first time — roughly 311,000 domains blocked between April and December 2020 — and, crucially, found about 41,000 of them appear to be collateral damage rather than deliberate targets. Earlier coverage had only shown the Firewall's blunt end, like the 2018 campaign that shut down 4,000+ websites and removed 147,000 pieces of content.
The measurement lineage continued after this paper: GFWeb later scaled the approach to track millions of domains and timestamped individual blocks like ChatGPT's, while follow-up work found users in Henan denied access to 5x more sites than a typical Chinese user — evidence that over-blocking varies by region, not just by domain. The 41,000 accidental blocks in this study are the first hint of that imprecision at national scale.
First-order effects
- Roughly 41,000 site operators — many outside China and never targeted — lose Chinese visitors without knowing why, since the Firewall gives no notice and the blocklist is opaque.
- Researchers and circumvention-tool builders gain the first quantified baseline of the blocklist's size and error rate, replacing anecdote with measurable scope.
Second-order effects
- Continuous monitoring projects like GFWeb emerge from this measurement tradition, turning one-off academic studies into running infrastructure that timestamps individual blocks and exposes the list's churn.
- The demonstrated over-blocking gives foreign site owners and CDN operators a concrete reason to treat China reachability as an engineering problem to monitor, not a binary blocked-or-open assumption.
Third-order effects
- Accidental blocks and system-wide failures — like the 74-minute block of all TCP port 443 HTTPS traffic in August 2025, possibly accidental — show the Firewall's scale makes precision impossible, meaning censorship error becomes a permanent, quantifiable feature rather than a bug to be fixed.
- If regional variance like Henan's holds as a pattern, enforcement is layered and uneven across provinces, complicating any single national picture of what Chinese users can actually reach.
The trend: Great Firewall censorship is shifting from an opaque black box to a continuously measured system, as academic tools expose both its scale and its structural over-blocking.