Access and hosting10 min readFacts checked on November 1, 2026

Is my website accessible in China?

One green result won't tell you. Here's how to test properly, without a VPN skewing the answer.

The short answer

Test your exact URLs from probes inside mainland China, across several carriers and at different times. Then check that the full page loads, not just the connection. Finish with a person on a normal Chinese network. A VPN, a Hong Kong test or a roaming phone doesn't count as proof.

  1. GreatFire's FAQ, read on November 1, 2026, says blocking is per address, so one article can be blocked while the home page loads.
  2. A 2025 study found a Henan provincial firewall blocking 4,196,532 domains in total, against 741,542 for the national firewall.
  3. On August 20, 2025, port 443 traffic between China and the rest of the world got injected resets for about 74 minutes, then it stopped.
  4. GFWeb (2024) saw HTTP requests for 7.8 million registrable domains redirected to Aliyun ICP warning pages, which it says isn't state censorship.
  5. Chinafy says 83% of its Beijing tests on 100 Shopify stores ran past 30 seconds or didn't finish, though GreatFire found shopify.com unblocked.

A site that opens for you can fail for a buyer in Shanghai, and nothing on your side will warn you. You can check most of it from your desk, as long as you know what each test really shows.

This guide is part of how to put your website in China. It covers one question: how to test your site, and how far to trust the answer.

How do you test a website from China?

Test exact URLs from probes inside the mainland, on several carriers, then check that the full page loads. Work in this order, because each step catches what the one before misses.

  1. List your exact URLs. Take your home page, your main Chinese page and your contact page. GreatFire says blocking is per address, and it tests http and https as separate addresses.1
  2. Check reachability. Run each URL through GreatFire or WebSitePulse. Both compare mainland results with a point outside the mainland.13 If GreatFire has never tested an address, its page offers a live test.1
  3. Spread across networks. Use BOCE, ITDog or 17CE to compare carriers and regions.456 On BOCE, pick your node, since a single check uses one and assigns it at random by default.4
  4. Check the whole page. Use a test that reports what loaded, such as AppInChina’s tester or Chinafy’s speed test.78 Compare it with a run from outside China.
  5. Repeat at another time. Try a different day and hour. A 2020 study found that cross-border slowdowns follow a daily pattern.16
  6. Ask a real person. Someone on a normal Chinese connection opens the pages and sends your form. See why forms fail in China for what to watch.
Fig. 1Order of tests
  1. Exact URLs
  2. Mainland reachability
  3. Carriers and regions
  4. Full page load
  5. Another day
  6. Real visitor

Which tool answers which question?

No single tool covers everything. Reachability tools tell you if a connection worked. Network tools show where it failed. Page tools show what actually loaded.

ToolBest forTests fromLimit to know
GreatFireBlocking of one exact addressSeveral mainland probes, plus an outside control 1Its headline figure covers 90 days, not this second 1
WebSitePulseQuick live checkShanghai, Beijing, Guangzhou, plus Hong Kong as a comparison 3No method published on the page 3
BOCEBlocking, DNS and carrier comparisonTelecom, Mobile and Unicom, data center or home nodes 4Chinese only, with a daily guest limit, no batch checks for guests, and no explanation of its detection method 4
ITDogPing, TCPing, HTTP, DNS and routeResults by node, region and carrier 5Described here by a third party 5
17CEFree performance testsAnalysis by region, carrier and test point 6Node count not stated on the page we read 6
GlobalpingDNS, HTTP, ping and tracerouteCommunity probes, selectable by country 1025Results shift with each probe’s provider 10, and the pages we read give no count of probes in China
AppInChina testerFull-page score from 0 to 100Servers it says are inside the mainland 7Vendor tool, and it says results vary by ISP and region 7
Chinafy toolsBlocked resources and load time3 servers in China among 18 9Vendor tool, and its code test covers a 10-page sample 8
ViewDNS and ComparitechQuick check of one domainViewDNS says “multiple locations” in the mainland, Comparitech names none 2627Neither page explains its method 2627
WordPress pluginsA check from your dashboardFive named mainland locations each 1112About one free test a month, and your URL goes to the vendor’s API 1112

We read each tool’s own page, or the source shown, on November 1, 2026. We didn’t run them side by side, and we found no independent comparison.

AppInChina calls its test free, and it covers DNS, HTTP, network diagnostics and browser rendering.723

Chinafy measured its Shopify benchmark with WebPageTest from Beijing, so that’s another option to check.17 Our speed guide covers what we could confirm about its agents.

Kinsta names four tools and seven test locations, and those are the same places WebSitePulse and Blocked in China name.20311 That adds to the doubt, but it doesn’t prove shared probes.

Node counts don’t settle it either. Juejin’s guide says the number of nodes isn’t the quality of the observation.5

Why one green result proves little

A pass describes one address, one network and one moment. Published measurements show how much can differ on each of those.

What can changeWhat the evidence saysWhat to do
The addressOne article can be blocked while the home page loads 1Test the pages that bring inquiries
The provinceHenan’s firewall blocked 4,196,532 domains in total, against 741,542 for the national one, from November 2023 to March 2025 with a 2024 gap, on traffic leaving the province 14Don’t let a Beijing pass speak for the country
The carrierGFWeb traced most gaps with other platforms to local ISP and cloud interference 13Compare Telecom, Mobile and Unicom
The momentOn August 20, 2025, port 443 traffic got injected resets for about 74 minutes, seen from a Beijing server, exact cause unconfirmed 15Repeat before you conclude
The directionAbout 1,000 domains trigger the HTTPS filter only from inside China 13Prefer probes inside, not outside

The Henan figures come from one carrier, China Unicom, and they stop in March 2025.14 The authors found no regional blocking in six other locations, though they say their vantage points were limited.14

Henan’s blocks were also short: 75% of blocked domains stayed blocked for under 51 days.14 We found nothing newer, so treat this as a known risk, not a current map.

Reachable doesn’t mean working

Shopify shows the gap. GreatFire recorded no blocking of https://shopify.com in 13 tests up to November 1, 2026.2

Yet Chinafy’s benchmark says 83% of its tests on 100 Shopify storefronts from Beijing ran past 30 seconds or didn’t finish. Only 50 to 54% succeeded from Singapore, Virginia and London.17 Chinafy sells the fix and gives no exact test date, so read those figures with care.

Both can be true. One tests an address, and the other tests a whole store with its images, scripts and fonts.

A fast result can even be a bad sign. Chinafy’s guide shows a page that looks quicker in China, and points to the number of resources loaded as the figure that matters.8

Its rule of thumb is that a gap above 5% in resources means parts of the page may display badly.8 That’s one vendor’s threshold, not a standard.

Why VPN, Hong Kong and roaming tests mislead

They don’t travel the network your visitors use, so a pass tells you little.

  • VPN. GreatFire lists VPNs among the connections that reach the internet differently from its probes.1 A pass may only show that your VPN works.
  • Hong Kong. WebSitePulse itself labels Hong Kong as “not behind the China firewall.”3 Chinafy, a vendor, says a clean test there “creates false confidence.”18
  • Corporate lines. GreatFire puts these in the same group,1 so a colleague in a multinational’s office may not be a fair test.
  • Foreign SIMs. UNSW’s Salil Kanhere told The Canberra Times that in many cases, a traveler’s data is routed through the home carrier or an overseas gateway, while a local Chinese SIM typically stays subject to local filtering.19 So don’t count a traveler’s phone as a test.
  • DNS lookups from abroad. Citizen Lab reported in 2021 that forged Chinese records had polluted public resolvers, including Google’s and Cloudflare’s.22 So a lookup outside China can mislead you too.

Free check

Is your website visible on Baidu?

Enter your website. We check Baidu, access from China and how your site is set up, then email you the report.

How do you read the results?

Read the pattern, not a single failure. Juejin’s guide suggests working through the layers in order: DNS, TCP, TLS, HTTP, then routing.5

What you seeLikely meaningNext step
Ping fails, but TCPing and HTTP workICMP is filtered, and your site may be fine 5Ignore the ping
Lost packets at a middle hop onlyNot proof of a problem 5Check the final result
One node failsPossibly local interference, not a national block 13Compare other carriers and regions
A warning page about ICPA Chinese cloud provider’s check, not the firewall 13If you host with that provider, check your filing with it. If not, retest from another network
DNS returns an IP that isn’t yoursPossible DNS poisoning 1021, or one of several IPs your domain normally resolves to 5Compare with your host’s or CDN’s IP ranges
Connection resets on TCP or TLS, on most nodes, while the outside control passesPossible filtering of your domain 13See if your IP or name is the cause
All mainland nodes fail, control passesCensorship, or a firewall or misconfigured server 10Check your host’s blocking rules first
Control fails tooInconclusive, and your server may be down 1Check the server, then retest
Everything connects, but slowlyA speed problemRead the speed guide
Connects, with fewer resources loadedBlocked third-party files 8Audit your Google services
GreatFire says “Not enough data”The test couldn’t tell, not that you’re fine 1Use another tool

The ICP row matters more than it looks. GFWeb saw Aliyun and QCloud redirect HTTP requests for 7.8 million and 6.8 million registrable domains to ICP warning pages. It says this isn’t state censorship, but hosting providers enforcing the ICP rules.13

For the full diagnosis, read is your site blocked or just slow.

What no test can tell you

A test shows what happened. It doesn’t tell you why, and it doesn’t tell you if search engines have your site.

  • Why a site was blocked. We found no way to learn the reason for a given site. Our guide to why websites get blocked covers the known causes.
  • Whether your IP is the cause. We found no dedicated tool for this. A URL test can’t always separate the IP from the name. See shared IP blocks.
  • Whether Baidu has your pages. That’s a separate check, covered in how to see which pages Baidu has indexed.
  • Whether it will last. A clean test today says nothing about next month, so build it into the China-ready checklist.

The bottom line

Test exact pages from inside the mainland, across carriers and on different days, then have a real person confirm it. Don’t count VPN, Hong Kong or roaming results.

Keep “the address connects” and “the page works” apart. Most of what looks like a contradiction between tools comes from mixing up those two questions.

How Index in China can help

We localize sites into Simplified Chinese, with a person reviewing every page, and get them indexed on Chinese search engines. Testing from China isn’t a listed part of that offer.

Being found only helps if your page loads, so bring your results when we talk. Book a 30-minute call.

Book a 30-minute call

Common questions

How often should I test my site from China?

We found no published standard, so let change decide. Test before launch, then after any hosting, CDN, DNS or plugin change, since each one can alter what your pages load. A regular check on the pages that bring inquiries also helps, because a test only describes the moment it ran.

Is there an official list of websites blocked in China?

We found none. The authors of the GFWeb study write that no official report on the firewall's design has ever been published. The domain lists you see in the press are rebuilt by researchers through measurement, and their totals differ by study, by method and by period, so don't add them together.

If my site passes, can Baidu index it?

Not necessarily. A test shows that a page opened for a probe. It doesn't show that Baidu crawled the page or kept it. Indexing is a separate check that you run on the search engine itself, and our Baidu guides explain how to do that.

Does my website platform change the result?

It can, so test your own domain rather than the platform's. GreatFire found https://shopify.com unblocked in 13 tests, while webflow.io was mixed across 13 tested URLs.224 Neither result describes your pages, because your theme, plugins and embeds load their own resources from other addresses.

Book a call
30 minutes

Find out what getting indexed in China would take.

In a 30-minute call we look at your site, your market and what getting indexed would take. You leave with a clear answer, whether or not we work together.

Book a 30-minute call
  1. 01Book a slot
  2. 02We look at your site
  3. 03You leave with a clear answer

Book a 30-minute call

30 minutes on your site, your market and what getting indexed would take.