Start your 3-day free trial
Sign up to experience all premium features at no cost.
*Available only to new users. Each user is limited to one trial.


Google, Facebook, YouTube and X are reported as blocked in China by the UK travel advisory, but a useful service list must distinguish that warning from dated network measurements and account rules. This full list covers the 15 services disclosed below, not every website on the Internet. A website result does not establish whether its app, login, calls or file transfers work.[1]
The OONI measurements below describe selected web addresses, not every function of an app.
Key Takeaways
- Every measurement row has a specified hostname, date window and network scope.
- Anomaly, confirmed blocking, test failure and successful access are different outcomes.
- Platform eligibility and app-store availability are separate from reachability.
- Prepare approved alternatives; do not depend on any single connection method.
Make a list of the actual tasks you need: contacting family, joining a work meeting, downloading a document or checking a booking. Ask your employer about approved connection arrangements and the correct service tenant. Save necessary travel information offline, and agree on a permitted backup communication channel. The China connectivity overview provides the broader planning context.
Where local law, network policy and platform terms permit your intended connection, AethoVPN can be prepared before travel for the network-access step: install the official Windows installer or Android APK, or obtain iPhone, iPad or Mac configuration through the official setup guide with Pro or Premium. Select a currently available location in the app and verify the exit with the IP-location tool before checking the specific permitted function. A successful tunnel does not guarantee that every listed service works or confer account eligibility.
Treat advance preparation as a chance to verify the permitted arrangement, not proof of future availability inside mainland China. If permission or legality is uncertain, obtain the appropriate advice instead of relying on a generic recommendation.
New users can start the 3-day Pro trial, available once per user.
The MIIT explanation concerns unapproved cross-border telecommunications business and describes lawful enterprise access arrangements. It should not be expanded into blanket permission for any personal VPN use. Our product paragraph is conditional; it is not a country-specific authorization opinion.[6]
The quantitative sample uses OONI's official aggregation API for country code CN and the Web Connectivity test, from 1 September 2026 inclusive to 1 October 2026 exclusive. It was retrieved on 5 October 2026. For each named hostname, we retained HTTPS root-page inputs with an empty path or /; we did not merge arbitrary subdomains or infer a service's entire infrastructure from its name.[2]
The aggregation's network field is the probe ASN, which identifies a network rather than a city or user. The count of ASNs shows the spread of this sample, not national representativeness. Volunteers choose what to measure; coverage is not a probability sample of China's residents, and repeated measurements are not independent users.[3]
N means total measurements, OK means successful test classification, A means anomaly, C means confirmed classification, and F means test failure. We retain all five rather than reporting a misleading “blocked percentage.” A failure is not automatically censorship, and an anomaly needs analysis before it becomes a confirmed conclusion.[3]
The source dataset is shared across all four language versions. No row is a test we ran from inside China. Retrieval date is not measurement date, and this September window does not prove the result on a later journey.
Each row covers the named root webpage, not app messaging or authenticated functions. Evidence date is 1–30 September 2026 for every row; checked date is 5 October 2026. The linked service article provides task context, not additional measurements silently added to these counts.[2]
| Category / service | Root hostname / function | N / OK / A / C / F | ASNs | Evidence conclusion |
|---|---|---|---|---|
| Communications / Telegram | telegram.org | 212 / 7 / 105 / 1 / 99 | 16 | Confirmed evidence in some measurements |
| Communications / WhatsApp | www.whatsapp.com | 209 / 18 / 99 / 0 / 92 | 15 | Mixed / anomalous; no confirmed result |
| Communications / Signal | signal.org | 204 / 8 / 94 / 1 / 101 | 16 | Confirmed evidence in some measurements |
| Social / Facebook | www.facebook.com | 705 / 22 / 137 / 0 / 546 | 18 | Mixed / anomalous; no confirmed result |
| Social / Instagram | www.instagram.com | 741 / 35 / 170 / 8 / 528 | 18 | Confirmed evidence in some measurements |
| Social / X | twitter.com | 712 / 21 / 148 / 0 / 543 | 18 | Mixed / anomalous; no confirmed result |
| Search / Google | www.google.com | 51 / 0 / 40 / 0 / 11 | 6 | Mixed / anomalous; no confirmed result |
| Work / Slack | slack.com | 46 / 46 / 0 / 0 / 0 | 7 | OK only in sample |
| Work / Teams | teams.microsoft.com | 47 / 45 / 1 / 0 / 1 | 6 | Mixed / anomalous; no confirmed result |
| Work / Dropbox | www.dropbox.com | 44 / 1 / 37 / 0 / 6 | 5 | Mixed / anomalous; no confirmed result |
| Video / YouTube | www.youtube.com | 700 / 21 / 143 / 0 / 536 | 17 | Mixed / anomalous; no confirmed result |
| Development / GitHub | github.com | 15 / 5 / 8 / 0 / 2 | 6 | Mixed / anomalous; no confirmed result |
| News / BBC | www.bbc.com | 62 / 1 / 27 / 1 / 33 | 12 | Confirmed evidence in some measurements |
| News / New York Times | www.nytimes.com | 71 / 1 / 47 / 0 / 23 | 9 | Mixed / anomalous; no confirmed result |
| Reference / Wikipedia | www.wikipedia.org | 44 / 1 / 40 / 0 / 3 | 6 | Mixed / anomalous; no confirmed result |
“Confirmed evidence” means some measurements received OONI's confirmed classification. It does not mean every measurement, ASN, function or later date was blocked. A mixed row retains its successful, anomalous and failed observations rather than forcing one countrywide label.[3]
For Google, Facebook, YouTube and X, the travel advisory separately warns of permanent blocking. That advisory is travel guidance and has no disclosed ASN-level measurement sample here. It supports its stated warning, not a rewriting of the OONI counts or a guarantee about every request.[1]
The Telegram, WhatsApp and Signal rows are webpage measurements. They do not exercise sending a message, establishing a call, fetching attachments or logging in. OONI has dedicated messaging tests, but those are not included in this article's Web Connectivity sample. Do not substitute a homepage result for an application test.[3]
The same boundary applies to office tools. Slack's root page had 46 OK classifications on seven ASNs in this window. That finding says nothing directly about a workspace's WebSocket connections or file uploads. Slack's official troubleshooting documentation identifies additional connection requirements, so the task needs its own checks.[4]
Teams had 45 OK, one anomaly and one failure across six ASNs. Microsoft also documents Teams operated by 21Vianet as a distinct China service context. A result for teams.microsoft.com is not a test of every 21Vianet endpoint, tenant relationship or meeting feature. Confirm with your administrator which service and account you actually use.[5]
GitHub's root-page sample is particularly small: 15 measurements across six ASNs. Even a successful page load would not verify clone operations, package downloads or Actions. The linked developer article addresses those tasks, while this list keeps the measured endpoint explicit.
All 15 rows have two further fields that remain unverified by this dataset: platform account or region eligibility, and availability in the reader's app store. We intentionally do not assign “available” or “unavailable” to either from a network result. A permitted account and a downloadable app require different evidence.
Google's Play guidance explains that account country or region affects which apps and content are available. That is a platform rule, not an OONI measurement. Do not assume changing the network address changes payment-country requirements, workplace membership or the account's supported region.[7]
Prepare software only through official or employer-approved channels and preserve recovery access permitted by your organization. Avoid unfamiliar installers offered as a “China version” without a verifiable publisher. The phone preparation guide covers the device-planning side without treating installation as proof of access.
If a platform says your account is not eligible, address that with the platform rather than recasting it as a network block. A transport change cannot supply missing contractual permission. Similarly, an app missing from a store does not alone show that its website is blocked.
Use uncertainty to choose a backup and a more specific question. For a mixed result, identify the hostname and function you need, the relevant network, and the date. A colleague's success on a different network is useful context, not a guarantee for yours. Avoid sharing credentials or private workspace names in a public report.
A successful test supports only the tested request. It may be useful when deciding which tasks need further checking, but it should not be advertised as universal reliability. An anomalous test is a lead for analysis; a failed test can also reflect circumstances unrelated to a block. None of these labels replaces the actual error details.
For practical fault isolation, use the mainland-China service diagnosis guide. That guide distinguishes maintenance and account issues; this article supplies a dated endpoint sample. Keeping those jobs separate avoids repeating a troubleshooting procedure inside every data row.
For travel, carry a connectivity preparation checklist and a permitted fallback. Do not repeatedly attempt sensitive logins during unexplained failures. If a work network or local rule forbids the proposed arrangement, stop and consult the responsible administrator.
No. It contains all 15 services in the disclosed sample. The Internet is larger, and measurement coverage does not support a claim of an exhaustive national list.
No. An anomaly signals a difference requiring analysis and can have other causes. Confirmed classifications have a narrower evidence basis; failures are retained separately rather than automatically called blocks.[3]
No. A webpage test does not exercise messaging, calls, authentication or file transfers. Those functions may use different endpoints and need appropriately scoped checks.
That claim is not supported by this sample. Their root pages had substantial OK results, while authenticated workplace functions and service tenants remain outside the tested scope.
No. Store availability, account eligibility and network reachability are separate. Installing an official app is preparation, not proof that every required function is available.
No. They cover September 2026 and the recorded networks, retrieved in October. They do not establish a later result on your particular network or device.
No. The permitted connection arrangement, network conditions, service policy and account requirements all matter. Keep approved alternatives and investigate the particular task rather than relying on a universal promise.
Disclaimer: VPN regulations vary by country and region and are subject to change. This article does not constitute legal advice. Please review and comply with your local laws before using a VPN.
Sources:
Sources checked 5 October 2026.
Related articles:
Sign up to experience all premium features at no cost.
*Available only to new users. Each user is limited to one trial.





