Can certificate registries be checked from inside China?

Mostly yes, on our test date: IAF CertSearch, the EU notified-body space, ILAC, CNAS English and NHTSA vPIC all returned 200 in three of three rounds from a mainland-China consumer connection. That matters because it is the working half of certificate verification: the China side can reach the international registries that name certificate holders.

Why can't automated tools check UL certifications?

On our test date UL's Product iQ database returned 403 to a scripted client on every round and every network path, even with a browser user-agent — the block keys on something deeper than the UA string. The practical consequence: a script or an AI agent cannot pull a UL listing for you; a human in a real browser session can. Checking whose name is on a UL 2849 or UL 2271 certificate is human work.

Does using a VPN or proxy help reach these registries?

Sometimes it does the opposite. In our measurement the two US government hosts — FCC and NHTSA vPIC — answered the direct connection and refused the proxy egress entirely, consistent with our earlier observations of government hosts rejecting commercial proxy IPs. A buyer or agent routing through a VPN may find exactly the official databases they need are the ones that stop answering.

How were these measurements taken?

Three GET requests per host per network path to the root or stated entry path, with a desktop-browser user-agent, on 9 August 2026 — direct from a mainland-China consumer connection and through a local proxy egress, with a Chinese commercial host and an international host as same-session controls. Every host returned the same status across its three rounds. No logins, no queries, no CAPTCHA interaction — front doors only, which is also the limit: a 200 on the root does not prove a search completes.

Measurement recordDual-path, same-session controls

“Check the certificate in the official registry.” We measured who can actually open the registries.

Every certificate guide on this site — and everywhere else — ends the same way: verify the holder’s name in the issuing body’s database. That advice silently assumes the database opens for you. We probed eight registries buyers are routinely sent to, three rounds each over two network paths, with controls in the same session. It mostly holds — with two exceptions that change how the checking has to be done.

Published 9 August 2026 · Measured the same day, front doors only

Prepared by Currawong Web’s China-side verification network.

1. The panel

Status codes, three rounds each, 9 August 2026. Every host returned the same status across its three rounds on both paths (one ISO round differed only between redirect and refusal).

Registry	What buyers use it for	Direct (from China)	Via proxy egress
IAF CertSearch	Management-system certificates (ISO 9001 family)	200	200
UL Product iQ	UL listings incl. UL 2849 / UL 2271	403	403
EU compliance space (webgate)	Notified bodies behind CE routes	200	200
FCC	Equipment authorization entry	403	no answer
ISO	Standard listings	403	302 / 403
ILAC	Accreditation-body recognition	200	200
CNAS (English)	Chinese accreditation body	200	200
NHTSA vPIC	VIN decoding, manufacturer records	200	no answer
Control: Chinese commercial host	—	200	200
Control: international host	—	no answer	200

The controls are what make the rows readable: the international control fails on the direct path (so the direct path really is an ordinary mainland connection) and succeeds via proxy (so the proxy path really is a working overseas egress).

2. Three findings

The China side can do this work. The registries that name certificate holders — IAF, the EU space, ILAC, CNAS, vPIC — all answered a mainland connection. When a certificate check or an ebike purchase needs a holder’s name confirmed, the databases are reachable from where the supplier documents are.
The hardest database is human-only. UL’s 403 was identical on both paths and unaffected by a browser user-agent — the detection keys on the client, not the address. A script, a crawler or an AI agent asking “is this UL number real?” gets a refusal; a person in a real browser session gets an answer. For lithium-battery goods, where UL 2849 / 2271 paperwork decides real safety exposure, automated verification is not on the menu.
Proxies can subtract access. Both US government hosts served the direct connection and dropped the proxy egress at the TLS layer — the same pattern we recorded for government hosts on the Chinese side. A buyer on a VPN, or an agent routed through commercial egress, may find precisely the official databases unreachable while everything else works.

3. Method, and what this panel cannot say

Front doors only. Three GETs per host per path to the root or stated entry path, desktop-browser user-agent, no logins, no queries, no CAPTCHA interaction. A 200 on the root does not prove a search completes; a 403 to our client does not preclude a browser session working — for UL that asymmetry is the finding.
One date, one connection per path. Availability is a dated fact: a host that refuses this week may serve next week, and a different ISP or egress may behave differently. Citing any row without its date misrepresents it.
“No answer” means no TLS-level response within 25 seconds across all rounds, distinct from an HTTP refusal — the distinction between a server saying no and nothing answering at all, kept deliberately, as in our earlier panels.

Citing this study

You are welcome to quote or reproduce these observations, including commercially, provided the measurement date (9 August 2026) and the stated limits travel with them. Machine-readable results: the full panel JSON. Related measurements: the Chinese official-source panel and the research index.