The most-blocked domains in the world
A single ranked list of what is being censored on the open internet right now. We pull every domain-tagged observation from the Voidly evidence table over the last 90 days and rank them by a composite importance score that captures three things at once: how many distinct countries are blocking the domain, how many blocked measurements we have (log-damped so one heavily-probed country cannot dominate), and how confident the underlying probes are in those flags on average.
Generated Thu, 21 May 2026 13:57:25 GMT · CC BY 4.0 · Raw JSON · Back to Atlas
Top 5 most-blocked globally
Ranked by importance score = n_countries_blocking × log(1 + n_blocked_obs) × avg_confidence. The flag strip shows the six countries with the most blocked observations for that domain in the window. Click a domain to see the full per- country breakdown.
3 ranked domains· MMED
| # | Domain | Category | Countries | Blocked obs | Block rate | Avg conf | Importance |
|---|---|---|---|---|---|---|---|
| 10 | youtube.com | MMED | 28 | 930 | 100% | 80% | 152.7 |
| 11 | instagram.com | MMED | 27 | 1,038 | 100% | 80% | 150.6 |
| 25 | tumblr.com | MMED | 23 | 416 | 100% | 90% | 124.9 |
Block rate counts blocked observations divided by total observations of the domain. For widely-probed domains (chatgpt.com, gemini.google.com) the global block rate is low even though the breadth is high — the importance score handles that by separately weighing breadth and volume.
Methodology
For every domain that appears in the Voidly evidence table over the last 90 days, we compute:
- n_countries_blocking — the number of distinct ISO-3166 alpha-2 country codes that have at least one blocked observation.
- n_blocked_obs — the count of evidence rows with signal type in dns-blocking, http-blocking-tcp-reset, http-blocking-timeout, block, interference, or http-blocking-blockpage.
- avg_confidence — the mean of the per-row confidence field across all blocked observations for the domain.
- recency_weighted_blocks — the sum of exp(-age_days / 30) over each blocked observation, so recent activity contributes more than archived rows.
The headline importance score is:
importance = n_countries_blocking
* log(1 + n_blocked_obs)
* avg_confidenceFilters: a domain must appear in at least 3 distinct countries and have at least 10 total observations (blocked or not) in the window. This drops one-off probes that would otherwise dominate the tail with noise.
Honest caveats. Per-row confidence values are produced by the upstream source (OONI annotates dns- blocking flags with ~0.7–0.9 confidence; CensoredPlanet Satellite flags have ~0.85–0.95). The category column is best-effort: most rows use Citizen Lab’s test list taxonomy where available and a curated map for well-known global-reach domains; everything else is marked OTHER. The window is rolling 90 days, so a domain that was heavily blocked 100 days ago can drop out of the ranking.