There is a companion finding on this site showing that the world’s most-blocked websites are gambling and pornography, not news. That is true — but it is the broad picture: lots of countries each block a handful of casino and adult domains as routine licensing or morality law. Ask a different question — not “what does the world block” but “what do the heaviest censors block” — and a second, opposite picture appears. The countries that censor deeply (hundreds of domains, not a dozen) overwhelmingly target political speech, not vice. The same dataset answers the two questions with two different villains.

Voidly’s /v1/measurement/intent-profile takes the domains a country blocks nationally — confirmed across at least three independent networks — and joins them to the 14,000-domain Citizen Lab category list, then reports three deliberately low-contestation focus shares: political speech (news, human rights, political criticism), regulated “morality” content (gambling, adult, drugs), and circumvention tooling (VPN/anonymity). Of the countries with a meaningful national-block list, here is where the depth sits.

The deepest censors, by what they block

Read it top to bottom: the two heaviest national censors in the corpus, Iran (789 domains) and Russia (608), are speech machines. Russia nationally blocks 264 news outlets — more news domains than the next-deepest country blocks of everything. Iran pairs 110 news sites with 81 human-rights domains. Then the depth flips: Indonesia and Thailand go just as deep but on gambling and adult content, with little news. Saudi Arabia and Turkey sit in between — genuinely mixed regimes that block news and vice. India’s list is the most spread of all (no single dominant category), which is itself the honest answer for India.

Why the global “gambling wins” ranking hides this

The most-blocked-domain leaderboard counts how many countries block each site, so it rewards breadth: betfair.com is blocked in ~10 countries because blocking gambling is legal and ordinary across a wide band of states. A single news site blocked by Iran counts once. So aggregate by domain and gambling wins; aggregate by how much one country blocks and the political censors tower over everyone. Both are true. They are not the same statistic, and conflating them is how a casino blocked in Belgium ends up in the same sentence as a news site blocked in Tehran.

Both poles are documented by the official record

The morality pole is almost arithmetical. Indonesia’s communications ministry (Kominfo) runs the TrustPositif filter, which has blocked over a million sites; of the 213,735 pages it ordered blocked in 2022, 156,975 were gambling and 51,588 were pornography — roughly 98% of the year’s blocking, with news and social a rounding error (OONI / iMAP 2022; Freedom House). That is the morality machine Voidly measures, in the censor’s own numbers. The speech pole is the mirror image: Iran and Russia rank among the world’s worst for press freedom and nationally block the news outlets themselves — RFE/RL, BBC, Deutsche Welle, Meduza and dozens more (Freedom House, Iran; Russia). Two well-documented censorship regimes, aimed at two different targets — exactly the split the confirmed-block data draws.

Honest caveats

This describes what is blocked, not intent in any legal sense, and the edges are soft. (1) Anomaly-based censors are under-counted. OONI records China’s Great Firewall interference as “anomaly,” not “confirmed,” so China shows only ~15 confirmed national-block domains here — nowhere near its true scope. The clean speech/morality split is most reliable for high-confirmation DNS-poisoning regimes (Iran, Russia). (2) Coverage of the tagged subset varies. Shares are computed over domains that carry a Citizen Lab tag (~86% of national-block domains overall, but only 24 of Italy’s 218 — Italy’s rotating piracy domains aren’t in the list — and small tagged counts for South Korea and Poland). An untagged domain is excluded from the shares, never assumed safe. (3) Free countries with tiny national-block lists (Netherlands, Spain, Canada here) are dominated by a handful of measurement false positives, not real policy, and are excluded from the thesis. (4) The confirmed-block layer is a periodic snapshot — see /v1/measurement/freshness for recency. Treat the focus shares as a qualitative fingerprint of a censor’s priorities, not a precise budget. Live: /v1/measurement/intent-profile.