Is your site blocked or just slow in China?

Blocked and slow look the same from a distance: someone in China says your site doesn’t open. They need opposite responses, so guessing wrong costs you time and money.

This guide lists what’s blocked, then shows how to tell which problem you have. It’s part of our work on putting your website in China.

What websites are blocked in China?

Google, the main Western social networks, Wikipedia and major Western news sites are blocked. We read GreatFire’s https test page for each address below on November 1, 2026, and added Wikipedia’s dates.

Address	GreatFire verdict	Failed tests, 90 days	Last test, 2026	Blocked since, per Wikipedia 15
www.google.com	Blocked 2	16 of 17	Oct 30	2014, fully
www.youtube.com	Blocked 3	4 of 4	Oct 20	2009
www.facebook.com	Blocked 4	6 of 6	Oct 1	July 2009 or earlier
www.instagram.com	Blocked 5	2 of 2	Oct 3	September 2014
x.com	Blocked 6	22 of 22	Nov 1	June 2009
www.wikipedia.org	Blocked 7	1 of 1	Aug 28	April 2019
www.reddit.com	Blocked 8	3 of 3	Oct 18	2018
www.nytimes.com	Blocked 9	4 of 4	Sep 12	2012
www.bbc.com	Blocked 10	3 of 3	Oct 12	October 2014
www.whatsapp.com	“Sometimes” 11	3 of 3	Sep 27	September 2017
github.com	“Sometimes” 12	42 of 63	Nov 1	Partial, since January 2013
www.linkedin.com	Not blocked 13	0 of 1	Oct 3	Partial access, per its table
shopify.com	Not blocked 14	0 of 13	Nov 1	Not in its table

Wikipedia’s table also carries an update banner from July 2025.¹⁵

Blocking also works per address. GreatFire says a home page may load while a specific article doesn’t.¹

GreatFire’s all-time list shows google.com at 40,316 of 41,317 blocked, without saying what it counts.¹⁶ Its domain summary rates 2,838 of 2,923 tested URLs as blocked.² Either way, not all of them. GreatFire’s blocked lists are the place to look for current addresses.¹⁶

For the services your pages load, see which Google services fail in China.

How many sites are blocked?

Nobody outside the system knows the exact number, but the studies below put it at a small share of the domains they tested. We found no official list, and every figure below was rebuilt from outside by researchers.

Study	Period	What it tested	Domains blocked
GFWatch 18	April to December 2020	DNS, about 411 million domains a day	311,000
GFWeb 17	February 2022 to September 2023	HTTP and HTTPS, 694 million domains	943,000 HTTP, 55,000 HTTPS
Wu et al. 19	November 2023 to March 2025, with a 2024 gap	HTTP and TLS, 227 million domains a week	741,542, cumulative

Don’t add these up. They cover different periods and methods. None of them counts IP blocking, so a shared IP problem wouldn’t show here.

Those are small shares of what was tested, but the test sets include many names nobody uses. Our guide to the Great Firewall works out the shares, and what they mean for your odds.

GFWeb could only categorize 79,500 of its base blocked domains, because many no longer existed or hosted no content.¹⁷

Business sites were among those it could classify, so a low rate isn’t immunity. Our guide to why websites get blocked in China covers the causes.

What’s the difference between blocked and slow?

A blocked site is stopped at the connection, so waiting doesn’t help. A slow site arrives in the end. There’s also a third state, and it’s the one people misread most.

• Blocked: the connection is forged, cut or dropped. A site can be fine everywhere else.
• Slow: every stage works, but the data takes a long time to arrive.
• Reachable but stuck: your server answers, yet the page hangs on something else it loads.

No official threshold separates slow from broken. Chinafy treats about 5 seconds as usable, which is a vendor’s working line, not a rule.²³

The third state has a clear mechanism. In one test from a mainland network, AppInChina, a vendor, saw a Google Analytics request time out after 3 minutes 9 seconds, and the page finished at 5 minutes 3 seconds.²¹

Cato Networks describes the same thing: pages that keep loading indefinitely.³⁰

Plain slowness is often normal. A 2020 study of more than 400 pairs of mainland and foreign nodes, over 53 days, found a bottleneck on 79% of pairs. More than 70% of pairs fell below 1 Mbps for over 5 hours a day.²⁰

In a one-day side test, bottleneck hops matched firewall hops in 34.45% of cases.²⁰ The authors call that a clue, since they only saw nodes that inject resets. So a slow site isn’t proof of censorship. Those tests date from 2019, and we found no newer independent measure.

How to tell which one you have

Find the stage where the request breaks. A 2026 Chinese developer guide says testing tools can split a request into DNS, connection, TLS, redirects, status and server response.²⁴

Use tools with probes on the mainland, as covered in how to test your site from China.

1. Check outside China first. If the control fails too, the problem may be the site itself, not censorship.
2. Test the exact address. Use the https:// address of a real page. GreatFire tests http and https separately, and the filters use different lists.¹¹⁷
3. Compare the IP address. Check that the DNS answer is yours, or your CDN’s. GFW Report documents forged answers, and they can pollute public resolvers outside China too.²⁵¹⁸
4. Test the port, not the ping. Run TCPing on port 443 against your real IP. A failed ping alone proves nothing, because ICMP can be refused while the site works.²⁴
5. Run an HTTP test. See which stage stalls: connection, TLS or server response.²⁴ If the status is normal and those stages are quick but the page is slow, look next at what the page loads.
6. Load it in a real browser. AppInChina says its tester loads the page in a real browser in China and reports which resources fail, slow down or block rendering.²⁸
7. Repeat across carriers and hours. Two or three samples aren’t enough.²⁴ Compare China Telecom, China Unicom and China Mobile, morning and evening.

1. 1Fails outside China tooSite problem
2. 2Forged IPBlocked
3. 3Connection cut or dropped on most mainland probes, works on controlLikely blocked
4. 4Connects, page hangsThird-party resource
5. 5All stages work, long totalSlow

Then read your results against this table. The readings are our inference from the cited research, not a published test standard.

What you see from mainland probes	Likely meaning
DNS returns an IP that isn’t yours or your CDN’s	DNS block 25
Connection is cut right after the TLS handshake starts	Name filter: the firewall reads the SNI and injects resets 17
TCP fails on port 443 on most probes, control fine	Possible IP block, including a shared IP 31
Document arrives, page hangs	Blocked or slow third-party resource 2130
Every stage works, total time is long	Slow: congestion on the way into China 20, or page weight
Worse at the same hours each day	Congestion: the 2020 study found stable daily patterns 20

GFWeb also found about 1,000 domains that trigger the HTTPS filter only from inside China.¹⁷ So only a mainland probe can clear that case.

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.

Your websiteCheck my site

Results that look like a block but aren’t

Several common signals don’t prove blocking. Check this list before you change host or call an adviser.

Signal	Why it misleads
Ping fails	ICMP may be refused while port 443 works 24
One test point fails	GFWeb traced most gaps with other platforms to local interference from ISPs and clouds, not the firewall 17
ICP warning page	Aliyun and QCloud devices redirected HTTP requests for 7.8M and 6.8M domains to these pages, even example.com on QCloud. GFWeb says this isn’t state censorship 17. By our reading, a probe inside those clouds can mislead
Everything fails for an hour	On August 20, 2025, all port 443 connections to and from China were disrupted for about 74 minutes, and ports 22, 80 and 8443 weren’t 26
A quick retry fails too	A first trigger can bring 90 seconds of resets, or up to 350 seconds of dropped packets for some domains 17
“Not enough data”	GreatFire says this means it couldn’t tell, not that the address is fine 1
Connection refused	GreatFire no longer counts refusals on otherwise clean routes as evidence of blocking 14

The ICP page has a real case too. Alibaba Cloud says its monitoring blocks a domain pointing to its mainland servers without a filing.²⁷ That’s a hosting issue, covered in ICP filing vs ICP license.

What to do with your verdict

Each verdict leads somewhere different, so don’t move hosts until you have one.

Verdict	Next step
Fails outside China too	Fix the site or server first
Slow	Work through how to speed up your website in China
Reachable but stuck	Replace what fails, starting with Google services
Possible shared IP block	Ask your host, and read the shared IP guide. Bluehost, for one, says it doesn’t offer a dedicated IP yet 31
ICP warning or host block	Review your hosting options with a local adviser
DNS or name block	Read what the Great Firewall means for your website, then talk to a local adviser

A new host or CDN can’t be assumed to help with a DNS or name block, because those filters act on the domain name, not the server.¹⁷ We found no published way to appeal one.

The bottom line

Don’t diagnose from a single test or a phone call. Find the stage that breaks, on several mainland networks, over several days.

A cut connection is a blocking question. A page that arrives late is a performance question. In both cases, the fix comes after the evidence.

How Index in China can help

We can’t lift a block, and we’ll tell you if your case needs a local adviser.

Our work is the next layer: we localize your site into Simplified Chinese, with a person reviewing every page, index it on Chinese search engines and send a monthly report on Chinese traffic. See what we do, or book a 30-minute call.

Book a 30-minute call

Monthly report BaiduIndexed Bing ChinaIndexed SogouIndexed