Method
How the entries were gathered, how they were checked, and how the list is kept current.
What was collected
General search engines, national portals, human-edited directories, vertical engines, metasearch, small-web indexes, archive gateways and link hubs. Services that merely syndicate a big engine's results were recorded rather than excluded, because where an index comes from is part of the map. Gone sites were kept as well.
How it was checked
- Every entry was fetched with
curl; the final URL after redirects and the<title>were kept. - The page body was read to decide whether the site is what it claims to be. Parked domains, businesses that took over a domain, and gambling pages were pulled out. A 200 proves a site is up, not that it is the one it was.
- 78 entries judged live were independently re-measured; 77 matched (the one exception had an expired certificate and was alive underneath). Entries judged gone were re-checked over a different route, and two misjudgements were corrected. The gone side is the one that needs the second look.
No URL from memory
Not one URL was listed without being fetched. A matching name proves nothing: Stract, widely listed as an independent-index engine, had a dead official domain, and the same-named domain belonged to an unrelated company.
Status words
| Status | Meaning | Link |
|---|---|---|
| Live | Reached on the measured date; content confirmed to match the name | yes |
| Degraded | Reachable but reduced, stale, expired certificate, or unstable from here | yes, with a note |
| Bot-gated | Cloudflare or similar blocked the machine. Not closed; visible in a real browser | yes |
| Hijacked | The domain is up, but it is a different site now | no |
| Parked | Parked or for sale | no |
| Gone | Unreachable or closed, confirmed over more than one route | no |
Observation point
Japan (JP). A single network, looking from the outside. Regional blocking means "unreachable from here", not dead. DNS absence, refused connections and timeouts were recorded separately; only what failed over several routes was marked gone.
Re-verification
Every entry is re-measured on a schedule. A hijack happens under HTTP 200, so changes in <title> and the appearance of gambling or for-sale words are routed to a human. Nothing is removed automatically; false positives have occurred. Only re-judged entries get a new date; untouched entries keep theirs.
Exclusions, paid listings, unknown indexes
Link lists for illegal streaming are not listed, however large they claim to be. Google and its country editions are out of scope. Directories that live on paid, reviewed listings are included in Tier 3; their order may be influenced by payment. Small directories rarely publish how their index is made, so their index is marked unknown rather than guessed.
Uneven coverage
Coverage differs by region. East Asia, the Russian-speaking web and Europe are dense; sub-Saharan Africa, Central Asia, the Pacific and the Caribbean are thin. Part of the reason is that the existing English-language lists show nothing there, and finding more needs research in local languages. Territories where nothing has been found yet are listed under Not yet covered (JA), and go on the map as soon as something is found.
Links and advertising
Every outbound link is nofollow. Hijacked, parked and gone entries carry no link; the generator refuses to emit one. There is no advertising.