Compare the national blocklists of every pair of countries Voidly measures and one pair stands far above the rest. Iran and Russia — the two heaviest speech-censors in the corpus — share 135 nationally-blocked domains (each confirmed across at least three independent networks). And on the politically meaningful slice the gap is starker still: the two share 23 of the same news and human-rights domains, where the next-closest country pair shares just five — a 4.6× gap on exactly the content that matters. One clarification up front: “convergence” here means the two blocklists independently land on the same targets, not that either copied the other. But what they land on, and how rarely anyone else blocks it, is the story.
Of the 135 shared domains, 23 are news or human-rights sites and the list reads like a directory of the international press-freedom and human-rights movement: amnesty.org (Amnesty International), cpj.org (Committee to Protect Journalists), article19.org, hrw.org (Human Rights Watch), freedomhouse.org, fidh.org (the International Federation for Human Rights), dw.com (Deutsche Welle), bbc.com, euronews.com. Both governments independently decided these specific organizations should be unreachable.
Another 25 shared domains are VPN and anti-censorship tools: expressvpn.com, cyberghostvpn.com, hide.me, hotspotshield.com, lantern.io (Lantern), getoutline.org (Outline), and bridges.torproject.org — the Tor bridge-distribution service specifically built to survive national blocking. Blocking the Tor bridges endpoint, not just torproject.org, is an unusually targeted move — though one any sufficiently determined censor can reach on its own.
The obvious objection: lots of countries block lots of popular sites, so overlap could be coincidence — the way many countries independently block betfair without sharing a list. The data rules that out here. At the confirmed national-block layer this analysis uses — a domain must be blocked across at least three independent networks before it counts — amnesty.org, cpj.org, article19.org, expressvpn.com and torproject.org each clear that bar in exactly TWO countries, and both times those two are Iran and Russia. These are not globally-blocked sites that any two censors would happen to share; they are rare, specific targets, and the same two governments reached them — convergence on the precise infrastructure of dissent and circumvention, not a coincidence of popularity.
One honest wrinkle, because it is checkable: Voidly’s looser per-domain
view (/data/domain/<d>) lists different, and often
more, countries for some of these domains — expressvpn.com, for instance,
shows up there under a handful of Western ISPs. That view counts any
single blocking ISP over a rolling 7-day window, including the transient
single-network false positives (a CDN hiccup, one flaky vantage point) that the
three-network confirmed gate exists to filter out. When the two layers disagree,
the confirmed layer is the rigorous one — and it is the one every number in
this finding is built on.
Both halves of the overlap are independently documented — which is exactly what “convergence, not coordination” predicts. On the Russian side, Roskomnadzor blocked Amnesty International’s Russian-language site in March 2022, alongside Human Rights Watch and Radio Free Europe/Radio Liberty, in the post-invasion media crackdown (Freedom House, Freedom on the Net 2024). On the Iranian side, DNS-tampering blocks of human-rights and women’s-rights sites are long documented (OONI, Freedom House). One honest nuance worth stating: an OONI study covering 2014–2017 found Amnesty’s site still reachable in Iran back then — so Iran’s block of it is a recent addition, which is precisely what Voidly captures: amnesty.org shows as a confirmed national block in Iran only in current (2026) data, not a decade ago. The two regimes arrived at the same target set from different directions and on different timelines — convergence, not a shared playbook.
(1) Convergence is not proof of coordination. The data shows the
two blocklists target the same things; it does not show that one copied the
other, that they share a vendor, or that they coordinate. Parallel, independent
targeting of the obvious infrastructure of dissent is the conservative
explanation, and it is enough to be alarming on its own. (2) “Blocked in
only two countries” means within Voidly’s measured corpus —
coverage is uneven, so a domain could be blocked elsewhere we do not densely
probe; read these as a floor, not a global census. (3) China is absent
from this comparison as a measurement artifact, not because it permits
Amnesty or VPNs — OONI records the Great Firewall as “anomaly,”
not “confirmed,” so China’s blocks rarely reach the confirmed
national-block layer this analysis uses. (4) The pairwise Jaccard score (0.107
for Iran-Russia) looks small only because each country’s full blocklist is
large; the signal is the political/tooling overlap and its rarity, not the raw
ratio. (5) The confirmed-block layer is a periodic snapshot — see
/v1/measurement/freshness. Live:
/v1/measurement/co-blocking.