Data quality & transparency
Good data.
Clear limits.
Check the evidence behind a conclusion.
See where measurement is strong or sparse.
Low confidence means limited evidence.
It does not mean a country is uncensored.
Scores unavailable.
The quality source could not be verified for this page load.
Check the sourceNo sample scores are substituted. The confidence score describes measurement coverage; it is not a censorship score or probability.
Country finder unavailable
Reload this page to try the source again, or open the raw quality API.
Open quality APIMeasurement context
Dates matter.
Gaps matter, too.
This separate summary describes the wider measurement corpus.
Freshest evidence
Nov 11, 2026Observation
Confirmed-block window
Nov 8, 2026From · 3d old at source generation
Check the source date before citing this window.
The freshest observation does not establish freshness for every country.
Domains with a category
66.2%- Categorized
- 3,763
- Uncategorized
- 1,918
Category comparisons cover the tagged subset. Untagged domains leave a gap.
Under review · 232 countries measured
Data as of 2026-11-08. Why this count is heldUnder review
Some blocked answers are on hold.
Three sources produced confident “blocked” answers that do not hold up. Until a fixed checker passes its review, every API route, MCP tool, page and dataset built on them answers “unknown” with the source’s data date, never “blocked”.
- The national blocklist layer. It let false positives through in open countries (one confirmed measurement could count a network), so it is being rebuilt with a stricter per-network rule. Its withheld 2026-09-20 to 2026-10-04 build listed facebook.com as nationally blocked in Germany and France.
- Platform, ISP and domain verdicts that counted CensoredPlanet rows and IODA outages as blocks. Neither is accepted as block evidence: CensoredPlanet rows include known false positives, and an outage is not a block. These verdicts listed Facebook as blocked in Singapore and the United Kingdom.
- The static ISP snapshot (ooni-isp-data.json, generated 2026-08-08). It marks a domain blocked when any one network lists it.
Held answers keep their non-verdict coverage fields, such as countries and networks measured. Country index scores, incident records and measurement freshness stay published with their own caveats. An answer of “unknown” is not a statement that a site is reachable.
Open to inspection.
How we measureHow to read the score
The published v1 methodology weights four measurement signals. These are score weights, not a forecast or a statistical confidence interval. The page displays the API’s scores without recalculating them.
- Freshness · 35%
- Most recent measurement: full weight within 24 hours, 0.7 within 7 days, 0.3 within 30 days, then zero.
- Volume · 30%
- Measurements in the last 30 days; the volume component saturates at 500.
- Stability · 20%
- Daily-count coefficient of variation (CV): 1.0 below 0.6, 0.6 below 1.2, and 0.2 for bursty sampling. Missing CV stays unavailable in the finder.
- Source diversity · 15%
- The number of represented feeds; the component saturates at three. A source count alone does not establish independent corroboration.
High confidence: 70–100. Medium: 40 to below 70. Low: below 40. These bands describe measurement support, not the severity of censorship. A high score does not guarantee a representative sample of every network or domain.
Measurement age is relative to the source snapshot. Source generation, block-window dates and page-check time are shown separately. A source timestamp more than two days old is flagged here for review; that flag does not change its published country score.
Coverage limits, sources and attribution
The quality endpoint and the measurement summary cover different cohorts and windows. Their country totals are not used to calculate a shared coverage percentage. Domain-category coverage above is calculated only from the summary’s measured and categorized domain counts.
The confirmed-block layer requires a domain to be blocked across at least three independent networks. It includes legal gambling and adult-content restrictions as well as other blocks. Anomaly-based censorship, including China’s Great Firewall, can be under-counted in that layer. An absent block record is not proof of accessibility.
Credit Voidly and include the endpoint, source timestamp and relevant limitations when using these scores. Upstream measurements retain their source licenses; read the source and licensing details before reuse.
Source methodology and missing data
Quality source
Source methodology unavailable.
Summary source
Source methodology unavailable.
Unavailable fields are never treated as zero. Invalid country scores are excluded from the spectrum; records with usable country codes remain in the finder. Conflicting bands are marked unavailable. Counts displayed above are calculated from the returned country records, with source-total mismatches disclosed.
Both endpoints are requested when this page loads. This view does not continuously refresh. Reload for another source check.
Page source check: . This is a request time, not an observation time.