CHINAGATEWAY.PRO

Article

How to Test Your Website from China Without Flying There

September 28, 2026

TL;DR

  • Four things to measure: first-render time from a mainland connection, blocked third-party dependencies, DNS resolution location, and whether your CDN edge sits inside the mainland.
  • Benchmarks: 15–30+ s loads and 2–8+ s TTFB are the global-only pattern; under 1 s TTFB is the tuned-mainland pattern.
  • Classify what you find: everything slow → transit problem; specific parts missing → blocked dependency.
Four measurements over a stylized map of mainland China — a stopwatch for first-render time, a dead dependency node, a DNS query path crossing the border, and a CDN edge node lit inside the mainland
Measurement beams from an offshore test point meet the lit mainland network — the stopwatches time every crossing

Your Shanghai colleague says the site is slow. That is a complaint, not a measurement. Before you open a budget conversation with your hosting provider or start rewriting DNS records, you need numbers — first-render time, blocked dependencies, DNS resolution path, and CDN edge location — collected from a mainland vantage point.

This article gives you the procedure to test your website from China — four tests, run from mainland China, classified into two failure modes, with benchmarks you can defend in a planning meeting. No flights required.

What You Are Testing For: Four Measurements

First-render time and TTFB from a mainland vantage

The first question is quantitative: how long does the page take to render for a visitor on a mainland connection? Open the page and record two numbers — the time until visible content appears (first render) and the time until the server sends its first byte (TTFB).

Global-only sites commonly take 15–30+ s to load in China. TTFB from China runs 2–8+ s on a global-only setup, compared with under 1 s with mainland hosting and tuning. If your numbers match the first pattern, the cause is transit and offshore origin — the cross-border hop adds latency to every request. For the underlying causes, see why your website is slow in China.

Blocked third-party dependencies

Open the browser's network tab and list every third-party domain the page calls. Note which requests time out or return errors. Google Maps, YouTube, and Vimeo embeds fail or time out for mainland visitors — the result is blank map tiles and dead video players. Font hosts, icon libraries, tag managers, and auth or payment APIs can all fail. A dependency audit maps which flows break. For the per-dependency replacement paths, see the blocked-API playbook.

DNS resolution location

Run a DNS lookup for your domain. Check where your nameservers resolve. Nameservers outside China add lookup round trips on every fresh connection. If the authoritative answer comes from a server in Virginia or Frankfurt, mainland resolvers must cross the border just to learn your IP address.

CDN edge presence

Check whether your CDN serves content from nodes inside mainland China. Offshore CDN edge nodes do not cut the cross-border hop. A CDN that terminates in Hong Kong, Tokyo, or Los Angeles still routes every request through an international transit link. The edge must be inside the mainland to remove that hop. What a mainland edge requires — and why the filing gates it — is the CDN glossary entry. Weighing a Hong Kong origin against a mainland one? See Hong Kong server vs mainland server — what changes and what does not.

Test Your Website from China Without a Mainland Office

The colleague method

The simplest vantage: a colleague on a mainland connection. Ask them to open your site, record the first-render time, and capture a HAR file or network-tab screenshot. This gives you a real-user data point from inside China. It is manual, but it costs nothing and it is real.

Mainland test-point services

Four services provide browser-accessible test points inside mainland China. Each lets you enter a URL and receive timing data from multiple vantage points.

boce.com offers a website speed test, Ping, DNS test, IPv6 test, ICP filing query, and Whois. Node coverage: "large numbers of high-performance nodes inside and outside China."

17ce.com provides real-time website speed tests, CDN tests, PING, DNS, HTTP, TRACEROUTE, and DIG. Monitoring points cover every domestic province and abroad, with line filters for China Telecom, China Unicom, China Mobile, Great Wall Broadband, and CERNET.

ping.chinaz.com lets you ping a server from multiple locations to check response speed. The page offers Telecom, Unicom, Mobile, and multi-line filters. Request-frequency limits are tiered across guest, member, and VIP access levels.

itdog.cn offers multi-location ping, tcping, website and API speed tests, traceroute, online MTR, and DNS queries. Its node pool includes paid residential home-broadband vantage points.

Verify your monitoring vendor's node list

Before you rely on a global monitoring vendor for China testing, check its public node list. Datadog, Uptrends, and Checkly all publish theirs, and none of the three lists includes a mainland-China vantage point — Checkly lists Hong Kong only. If mainland China is absent from the list, the tool cannot measure what your mainland visitors experience.

Testing whether your site is blocked

GreatFire, an organization that has fought censorship since 2011, uses a simple methodology: compare access from outside China against access from inside. If the site loads from outside but fails from inside, it is labeled blocked. This is a binary test — "is the domain unreachable?" — not a latency diagnosis.

ISP-line variance

China's three major carriers — Telecom, Unicom, and Mobile — route traffic differently. A site that loads acceptably on one carrier may time out on another. When your test tool offers carrier-line filters, run the test on each line and record results separately. Do not average across carriers.

What to record per run

For each test run, record:

Reading Your Results: Blocked or Just Slow?

Two failure classes cover most findings.

Everything is slow. The page loads, but takes 15–30+ seconds. All requests eventually complete — just late. This is a transit problem: your origin sits offshore, and every byte crosses the international link.

Specific parts are missing. The page renders partially — blank map tiles, dead video players, system-default fonts. Some third-party calls never complete. This is a blocked-dependency problem. The domain is unreachable from mainland China regardless of origin speed.

Benchmark table

MetricGlobal-only patternMainland-hosted pattern
First-render time15–30+ sImproves with mainland hosting and tuning — measure your own baseline
TTFB2–8+ sUnder 1 s
Third-party dependenciesSome time out (maps, video, fonts)Blocked ones replaced or proxied
DNS resolutionNameservers outside ChinaNameservers inside China
CDN edgeOffshore nodes onlyMainland edge nodes present

From Symptom to Fix: A Decision Matrix

The four measurements produce findings. The matrix below maps eight common symptoms to the likely cause, the test that confirms it, and the fix path. Every cell traces back to the measurement procedure above — no new numbers, just the routing logic.

Symptom you observeLikely causeTest that confirms itFix path
Everything loads — just late: 15–30+ s full render, TTFB 2–8+ sTransit: offshore origin; every byte crosses the international linkFirst-render + TTFB from a mainland vantage, per carrier line, never averagedMainland hosting + a CDN that terminates inside China; read the cost frame first; then the infrastructure assessment
Server answers fast, but the page still crawls or parts hangBlocked third-party dependency — some calls never completeNetwork tab: list every third-party domain; mark timeouts and errorsDependency audit → replace, proxy, or self-host per dependency; the per-dependency playbook carries the replacement paths
Blank map tiles, dead video players, system-default fontsSpecific embeds blocked (Google Maps / YouTube / Vimeo class)Network tab: which embed/SDK domains never complete from the mainland sessionReplace per dependency: maps → Amap/Baidu, video → domestic host; the blocked-API playbook lists each swap
Fresh visits slow; repeat visits somewhat betterDNS resolution outside China — lookup round trips on every fresh connectionDNS lookup: where do the authoritative nameservers answer fromMove DNS resolution inside China; no filing attached to this move; the slow-in-China playbook covers the DNS row
“Our global CDN covers China” — but mainland loads stay slowOffshore CDN edge: a Hong Kong, Tokyo, or Los Angeles node still crosses international transitCDN edge check: which node serves mainland visitorsMainland-terminating CDN from a licensed operator; filing per acceleration domain, unfiled domains blocked at the provider; the CDN glossary entry explains the filing gate
Fine on one carrier, times out on anotherCarrier-line routing variance (Telecom / Unicom / Mobile)Run each carrier line separately via the tools’ line filters; record per lineTreat per-carrier results as separate evidence in tuning and provider decisions; the CN2/CMI glossary entry covers the per-carrier doctrine
A colleague says “slow” — no numbers attachedA complaint, not a measurementColleague method: first-render time + HAR file or network-tab screenshot; full per-run record listBring the numbers and classify against the two patterns — everything slow vs. specific parts missing
You need a defensible “is it blocked at all?” answerReachability question, distinct from latencyOutside-vs-inside comparison (GreatFire-style binary)Blocked → dependency path; reachable-but-slow → transit path

Note: one run is a snapshot. Re-run the four tests after every infrastructure change, new third-party integration, or DNS modification. Map findings to the first six items of the 10-point China readiness diagnostic.

What to Do with the Results

Transit findings → mainland hosting. If your numbers match the global-only pattern, the next step is an infrastructure assessment. Mainland hosting on Alibaba Cloud or Tencent Cloud, with a CDN that terminates inside China, removes the cross-border hop. What a mainland CDN requires first is its own checklist. Our infrastructure service covers the architecture.

Blocked findings → dependency audit. If specific domains time out, identify every external script, API, font, and embed your site calls. Map which ones fail behind the firewall. Replace them with mainland-accessible alternatives, proxy them through a mainland-reachable intermediary, or self-host the assets. The infrastructure service includes dependency audit.

When the timed-out list includes payment gateways, that finding usually becomes a payments decision of its own. See WeChat Pay for foreign companies for the acceptance paths.

Compliance inventory → ICP filing check. If mainland hosting is indicated, check who holds the ICP filing. A mainland-hosted site requires an ICP filing (Bei'an) on a mainland-registered entity. Filing typically takes 2–4 weeks. See the ICP filing timeline for the process. Our ICP & compliance service runs the filing end to end.

Make It Repeatable

One test run is a snapshot. A measurement cadence catches regressions. Re-run the four tests after every infrastructure change, new third-party integration, or DNS modification.

Map your findings to the first six items of the 10-point China readiness diagnostic:

  1. Does your site take 15 seconds or more to first render from a mainland connection?
  2. Is the origin your China visitors hit hosted outside mainland China?
  3. Does your CDN serve China visitors from edge nodes outside the mainland?
  4. Do your domains resolve through nameservers outside China?
  5. Do you embed Google Maps, YouTube, or Vimeo?
  6. Do login, checkout, chat, fonts, or tag-manager flows call endpoints that time out from China?

If items 1–6 return "yes," your visitors have a materially worse experience than your analytics suggest. Items 9–10 on the diagnostic are launch gates — ICP filing and entity requirements — and should be addressed in parallel.

FAQ: Testing your website from China

How can I test my site from China without being there?

Ask a colleague on a mainland connection to open your site and record first-render time and network-tab data. Or use a mainland test-point service — boce.com, 17ce.com, ping.chinaz.com, or itdog.cn — that runs tests from nodes inside China. Record TTFB, timed-out domains, DNS resolver location, and CDN edge per run.

Is my site blocked in China or just slow?

Classify by symptom. Everything slow (15–30+ s) = transit problem — offshore origin, cross-border latency. Specific parts missing (blank tiles, dead players, timed-out fonts) = blocked dependency. Different failure modes, different fixes.

What load times should I expect?

Global-only: 15–30+ s load, 2–8+ s TTFB. Mainland-hosted with tuning: under 1 s TTFB. If your numbers match the first pattern, the cross-border hop is the bottleneck.

What should I test besides speed?

Three additional dimensions: blocked third-party dependencies (maps, video, fonts, tag managers, auth/payment APIs), DNS resolution location (nameservers inside or outside China), and CDN edge presence (mainland nodes or not). All three affect the visitor experience independently of raw load time.

Can Datadog, Uptrends, or Checkly measure my site from inside China?

No. All three publish their test-node lists, and none includes a mainland-China vantage point — Checkly lists Hong Kong only. If mainland China is absent from the list, the tool cannot measure what your mainland visitors experience. Always check the vendor’s public node list before trusting its China numbers.

Which mainland test-point service should I use?

All four work browser-side: boce.com (speed, Ping, DNS, IPv6, ICP filing query, Whois), 17ce.com (speed, CDN, PING, DNS, HTTP, TRACEROUTE, DIG, with Telecom/Unicom/Mobile/Great Wall Broadband/CERNET line filters), ping.chinaz.com (multi-location ping with carrier filters and tiered frequency limits), and itdog.cn (tcping, traceroute, MTR, DNS, paid residential nodes). Pick by the measurement you need; run carrier lines separately either way.

Why do my China test results differ between runs and between carriers?

Conditions change over time and China’s three major carriers route traffic differently — a site that loads acceptably on one carrier may time out on another. Record a timestamp and network type with every run, test each carrier line separately, and never average across carriers.

What exactly should I ask my mainland colleague to capture?

Three things: the first-render time, a HAR file or network-tab screenshot, and the per-run record — timestamp, network type, TTFB, timed-out domains, DNS resolution location, and CDN edge location. That is the minimum a planning meeting can defend.

How do I prove my site is blocked rather than just slow?

Compare access from outside China against access from inside — the GreatFire-style binary test. If it loads outside but fails inside, the domain is blocked. Then run the network-tab pass from a mainland vantage to see whether anything else is degraded alongside.

My TTFB is fast but the page still loads slowly — what does that mean?

Fast first byte with slow full render points past the origin: the usual cause is third-party calls that never complete — fonts, tag managers, maps, video, auth or payment APIs. List every timed-out third-party domain in the network tab and classify each: works, slow, or blocked. That list is your dependency audit input.

How do I check whether my DNS resolves from inside or outside China?

Run a DNS lookup for your domain and check where the authoritative nameservers answer from. Nameservers outside China add lookup round trips on every fresh connection — if the answer comes from Virginia or Frankfurt, mainland resolvers cross the border just to learn your IP.

Our CDN advertises global coverage including China — will the edge test still fail?

Probably, and the test is right. Offshore edge nodes do not cut the cross-border hop: a CDN that terminates in Hong Kong, Tokyo, or Los Angeles still routes every request through international transit. A mainland edge requires a licensed operator, and each acceleration domain needs a filing — unfiled domains get blocked at the provider.

How often should I re-run the four measurements?

After every infrastructure change, every new third-party integration, and every DNS modification. One test run is a snapshot; a cadence catches regressions. Map each round of findings to the first six items of the 10-point China readiness diagnostic.

Next steps

Testing tells you what is wrong. Fixing it requires the right infrastructure, the right dependency choices, and the right compliance posture.

Why Your Website Is Slow in China (and What Fixes It)

The three compounding causes your four measurements will confirm.

Why Your Site Breaks in China: The Blocked-API Playbook

Classify each timed-out domain: works, slow, or blocked.

China CDN for Foreign Companies: What It Requires First

Verify CDN edge presence — and the filing behind every mainland edge.

Have your test results?

Send them over — we will tell you which findings are fixable in weeks and which require a structural change. Start with the infrastructure service or the 10-point China readiness diagnostic.

hello@chinagateway.pro