What a guide is built on
Each site guide rests on a fact dossier assembled before writing: live reachability of the target, its official domains, apps and onion address, the OONI country aggregation for the last 90 days, and the legal context of its category. The article is generated from that dossier by deterministic templates - no language model adds facts - and every sentence is checkable against it.
Gates before publishing
- every number in an article must exist in the dossier or be derived arithmetically from it
- every domain mentioned must be in the dossier, so no invented mirror can slip in
- the category's legal note is mandatory
- a shingle-similarity gate rejects articles too close to anything already published
- a site with no measurement data at all is written as "no data", not as "probably fine"
Evidence classes
We separate what we measured (OONI probes, our own reachability checks), what someone else documented (court orders, regulator lists), and what the operator claims (mirrors, app links). The article says which is which. Block types follow from the evidence: a block page in the data means HTTP or DNS interception, silent connection failures mean SNI or IP filtering.
Update cycle
Reachability runs daily. Published articles re-verify on a 90-day cycle, and any status change (site closed, site back, blocking data appearing) re-writes the affected guide. Legal notes are reviewed when a country changes its rules.
Corrections and limits
We can be wrong: OONI probes are sparse for small sites, and a low count means few probes, not no block. Write to websiteunblock@my.ali.as with evidence - corrections are published. Nothing here is legal advice.