Internet censorship by country in 2026: a map and monthly index from OONI data

This index shows where internet blocking appears in OONI data for September 2026. We aggregate 4,013,093 measurements from OONI across a fixed panel of 25 websites. We sort every country and website pair into confirmed blocking, signs of possible blocking, no sign in the data, or not measured.

What OONI measures, and what counts as blocking

OONI measures whether a website loads from the network where a volunteer runs its OONI Probe app, and compares the result with a control. OONI is a non-profit free software project founded in 2012, and it publishes the results as open data.

This index uses OONI's Web Connectivity test, which checks websites from OONI's test lists. It runs a DNS lookup, a TCP connection and an HTTP request from the volunteer's network and from a control server, then compares the two. If the results match, there is no clear sign of interference, and if they differ, OONI flags an anomaly. A test ends in one of four outcomes, and the table shows how each is read.

Source: OONI FAQ, glossary and Web Connectivity description. Checked 14 October 2026.

A confirmed block and an anomaly are never added together, and an anomaly is never counted as a block.

Which countries have confirmed blocking?

The map and ranking below show the countries with confirmed blocking at national scale in the September 2026 data. Only 5 countries show confirmed blocking at national scale: Iran (21), Russia (12), Türkiye (1), India (1) and Myanmar (1). Iran has the most confirmed blocking: 21 of the 24 websites in our panel are blocked there at national scale.

National scale means a website is confirmed blocked on at least 2% of its tests in a country, from at least 30 tests. It must still reach 2% when the single biggest network is left out. Widespread means at least 20% of tests, and still 20% or more without the biggest network. Partial means 2% or more but under 20% once the biggest network is left out, so some networks may not block the website at all.

We rank by the number of websites blocked at national scale, because a count of blocked tests means little without the tests behind it. The panel's 25 websites include 9 VPN provider sites and the Tor Project site, so some of the counts below are VPN and Tor sites.

Why only 5 countries? Three things keep a country off the list:

• OONI confirms a block only when it sees a block page or a known censorship address, so blocking by DNS tampering or connection resets, as in China, appears as anomalies.
• 118 countries and territories have too little data to judge or no OONI record at all, and the map hatches them.
• 17 country and website pairs were removed because one network did the filtering.

The list shows confirmed blocking in OONI's data. It does not rank how censored each country is, and it does not say the other countries are open.

What changed since the previous edition

The table lists confirmed moves in pairs measured in both months. Changes in signs of possible blocking are listed separately and are not blocks. Pairs that gained or lost enough tests to be classified are changes in measurement volume, not in blocking.

Which countries block VPN websites?

The table shows VPN provider websites with confirmed blocking at national scale, country by country, with the share of tests that were confirmed blocks. VPN provider websites are blocked at national scale in 2 countries: Iran (8 of 8 sites measured) and Russia (7 of 8 sites measured, on some networks). The table covers each provider's website, not its app.

Scroll sideways to see every column

Country	ExpressVPN	NordVPN	Surfshark	Proton VPN	PureVPN	CyberGhost	Private Internet Access	IPVanish	TotalVPNno OONI data
IranIR	94.0%188 of 200Widespread †3 or fewer networks with enough tests	93.0%198 of 213Widespread †3 or fewer networks with enough tests	90.5%191 of 211Widespread †3 or fewer networks with enough tests	90.3%195 of 216Widespread †3 or fewer networks with enough tests	92.3%169 of 183Widespread †3 or fewer networks with enough tests	95.3%201 of 211Widespread †3 or fewer networks with enough tests	95.6%173 of 181Widespread †3 or fewer networks with enough tests	96.9%186 of 192Widespread †3 or fewer networks with enough tests	Not measured
RussiaRU	25.2%316 of 1,254Partial	18.6%336 of 1,809Partial	22.3%319 of 1,429Partial	21.5%343 of 1,597Partial	26.9%303 of 1,126Partial	24.3%305 of 1,255Partial	26.4%302 of 1,144Partial	No sign	Not measured
Countries blocking it at national scale	2	2	2	2	2	2	2	1	0
Countries with enough measurements	56	69	67	70	50	58	51	51	0

• Widespread confirmed on at least 20% of tests, and still at 20% or more without the biggest network
• Partial confirmed on 2% or more of tests, but under 20% once the biggest network is left out
• One network only not counted
• Anomalies possible blocking, not confirmed
• No sign no sign of blocking in the tests we have, or a few isolated confirmed tests
• Not measured fewer than 30 measurements

† 3 or fewer networks with enough tests

Not on this list does not mean open: Myanmar, Kazakhstan, Türkiye, Venezuela, Pakistan, Switzerland, Denmark, Kenya, Puerto Rico and China show signs of possible blocking on VPN sites, which OONI has not confirmed. Anomalies on VPN sites also show up in countries with no confirmed blocking, so treat them as a reason to look, not as blocks.

• Iran: 8 of 8 VPN sites measured are blocked at national scale, at confirmed rates of 90.3% to 96.9%. Per site, blocking shows on 2 of the 2 networks with enough data. On 8 of them one network, AS42337 (Respina Networks & Beyond PJSC), holds 52.2% to 61% of the confirmed blocks; on all other networks the rate is 80.9% to 93.1%.
• Russia: 7 of 8 VPN sites measured are blocked at national scale, at confirmed rates of 18.6% to 26.9%. Per site, blocking shows on 1 to 2 of the 7 to 15 networks with enough data. On 7 of them one network, AS39028 (JSC "ER-Telecom Holding"), holds 77% to 86.8% of the confirmed blocks; on all other networks the rate is 4.1% to 5.9%, so 6 of them read as partial and not widespread.

Source: OONI web_connectivity measurements for September 2026, aggregated and classified by WhatWorks.tech. Only countries with a VPN site blocked at national scale are in the table. Provider names link to our profiles where we have one.

This matters if you may need a VPN while you travel. If a provider's website is blocked where you are, signing up and installing its app may be harder. Two providers say what to do about it:

• Surfshark advises installing and setting up its app before you board the plane, because in some countries an app store can be out of reach. Our Surfshark profile covers its TV apps.
• Proton VPN says its mobile apps are also available from the Apple App Store, Google Play, the Amazon app store and F-Droid when its website cannot be reached. Our Proton VPN profile covers its TV apps.

A blocked website does not prove that the app is blocked, and an open website does not prove that the app connects. OONI's test loads the website only, so these rates say nothing about whether a VPN connection works in a country. Once you are connected, our VPN check shows which country websites see. Provider names in the table link to our profile when we have one.

Iran and Russia: how confirmed blocking shows up

OONI's own descriptions explain why the two countries look different. OONI lists Iran among the countries where ISPs serve block pages, a form of blocking it confirms automatically. For Russia, OONI reported in 2022 that every ISP implements government-mandated blocks independently, so blocks are not seen on all networks and the techniques differ.

Iran

The Iran spotlight lists each website with confirmed blocking, its rate, and how many networks with enough data block it on their own.

Russia: blocking that varies by network

OONI's 2022 report describes Russian blocking as differing from ISP to ISP, so one network can carry much of the confirmed blocking. We check that, and we describe such cases as blocking on some networks, not as nationwide. The section on one-network cases below lists the pairs where one network still carries most of the blocking.

China and other signs of possible blocking

China is a clear case of blocking that OONI mostly cannot confirm. In 2023 OONI found its own website blocked in China in two ways. DNS injection sends forged answers to website-address lookups, and connection resets during the TLS handshake cut the connection off while it is being set up.

OONI called the injected DNS answers typical of the Great Firewall. Neither technique produces a block page, so under OONI's current rules this kind of blocking mostly shows up as anomalies. OONI also says China has one of the most advanced and pervasive levels of internet censorship in the world. A confirmed-only count would make China look freer than it is, so we list its anomalies separately and never add them to the ranking.

In 2023 OONI reported that China appeared to be stopping OONI Probe users from submitting measurements. We do not know whether that explains the numbers today.

In September 2026, China has 21 websites above the anomaly threshold of 20%, 14 of them outside VPN and Tor sites, from 5,604 measurements in total.

Other countries with signs of possible blocking

The list below shows every country with at least one website above the anomaly threshold, with VPN and Tor sites counted apart. It is not a ranking of censorship and not a list of blocks, and we never add anomaly counts together.

In August 2025 OONI reported that its data suggests X has been blocked on some networks in Tanzania since May 2025. OONI says the block is not seen on all tested networks, and the techniques include DNS hijacking, TLS interference and IP-level blocks. The report covers X only. In this edition, Tanzania has no website at national scale and signs of possible blocking on 5 websites, including Twitter (twitter.com), X (x.com) and Telegram.

VPN and Tor sites need a closer look. 5 of the 24 countries with any signs of possible blocking show them only on VPN and Tor sites: Switzerland, Türkiye, Kazakhstan, Denmark and Kenya. These sites also show anomalies in countries with no confirmed blocking, so read them with extra care.

OONI warns that an anomaly may be a false positive. Its documentation names an unstable network, an unreliable server, a DNS resolver that returns a nearby address and a website that serves different content by country. A website owner who blocks visitors from a country looks the same as censorship in the data. We call a website blocked at national scale only where confirmed blocking holds without the biggest network.

Why one network can make a country look blocked

One network that filters content can make a whole country look blocked when many tests run on it. OONI records the network of every test by its ASN, a number that identifies the network a test ran on, and an ISP usually has at least one. Censorship can differ from network to network within a country. Two checks in this index guard against that:

• The single-network rule looks at every pair that would count at national scale. We pull the per-network breakdown, and the rule applies when one network holds at least 50% of the confirmed blocks. If the other networks together, from at least 30 tests, show a confirmed rate under 2%, the pair is labeled one network only and left out of the ranking.
• A second screen flags a country for a closer look when its national-scale candidates, before the network check, are only VPN and Tor sites, with none widespread.

The flagged pairs for this edition are below, with the network's ASN and the name OONI reports for it. The single-network rule moved 17 country and website pairs that would otherwise have looked like national blocking into one-network cases, in Finland (12), Mexico (3), Myanmar (1) and Thailand (1). We do not know what kind of organization each network is.

The rule removes only the biggest network. A pair can stay at national scale while one network carries half or more of the confirmed blocking, as long as the rest of the country still shows blocking. A widespread pair like that is relabeled partial when the rate without that network is between the partial and widespread thresholds. The pairs below stay in the ranking, with the evidence.

The rule has limits. Two filtering networks can still pass, and it cannot see blocking that never produces a confirmed block. We show how many networks with enough data block on their own, and we describe pairs with few such networks as blocking on some networks.

Full country-by-website matrix and data gaps

The matrix shows every country with OONI data against every website, with one status per pair. Isolated confirmed blocks, marked i, are a few confirmed tests under 2% of a pair's tests, and they are not counted.

Not measured means too few tests to classify in the window, and it is not the same as no sign of blocking. 37.3% of the country and website pairs with any measurement have fewer than 30 measurements, so they are not measured. OONI tests a sample of websites from curated test lists, and only where volunteers run probes, so coverage varies from month to month. TotalVPN returned no OONI measurements at all in this window, so it is shown as not measured.

Thin coverage cannot show that a country is uncensored. OONI says a confirmed block appears in its data only if the website was tested on the specific network that blocks it and the measurement was published.

How the index is built, and its limits

Each edition covers one calendar month of OONI Web Connectivity measurements. The panel is a fixed set of 25 websites: the websites of social platforms and messaging services, news and video sites, reference sites, circumvention tools and VPN providers. OONI counts VPNs and proxies among the common circumvention tools. We publish each edition after the month closes and the data is checked, usually within two weeks, and the current edition is September 2026.

1. We read OONI's aggregation API, which returns counts of measurements and of confirmed, anomalous, failed and OK results by country, network and website.
2. We classify every country and website pair with the thresholds below, using only pairs with at least 30 measurements.
3. For each pair that would count at national scale, we pull the per-network breakdown and apply the single-network rule.
4. We rank countries only by the number of websites blocked at national scale.

These limits apply to every figure on this page:

• This index covers blocked websites only. It does not count internet shutdowns, slowdowns or app-store removals.
• A country with no OONI measurements in the window is not in the tables, and the map hatches it.
• Volunteers choose where and when to run tests, and they can opt out of sending measurements, so coverage is uneven.
• Confirmed blocking undercounts blocking that leaves no block page, as in China.
• Each test loads a website. It does not test apps, and it does not show what a person can do after connecting.
• The panel is a fixed set of websites from OONI's test lists, not the whole web.
• OONI can add late measurements after we pull a window, so a re-run can move a count by a few tests.
• The thresholds are our choices, shown in the table above, and a different line would move some pairs.

Data on request, license and how to cite

The country and website data behind this page is available on request for research and journalism. Email [email protected] and say what you plan to use it for. To inspect individual measurements yourself, use OONI Explorer.

OONI publishes its data under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 license, which covers all published OONI data. Our classified data is derived from it and carries the same license.

If you reuse or cite it, credit OONI by name, say that WhatWorks.tech aggregated and classified the measurements, and keep any reuse non-commercial and share-alike. Read the OONI data license first. A suggested citation line, with the edition month, is below.

Related VPN profiles and tools

• Our VPN provider profiles: how the VPNs in the table compare and what each one offers.
• Our VPN IP checker: whether an IP address looks like a VPN, a data center or a Tor exit, and which network owns it.
• What is my IP: the country and network your connection reports, the same kind of identifier OONI records.