A free proxy list by country filters our live pool down to one exit location. At 10:59 UTC on 28 September 2026 we read the whole pool in one query and counted it by country. The same day, our server in the United States tested free proxies from the 34 countries with the most real supply.
The counts hold two surprises. First, most of the list sits on a few dozen addresses. Second, the country with the most real addresses is not the one most lists put first.
Most of the list sits on 44 addresses
Our pool held 16,749 live entries on 3,862 addresses in 112 countries. Yet 12,746 of those entries, 76.1 percent, sat on just 44 addresses. Each of those answers on 20 ports or more, and every port counts as an entry.
Thirty-nine of the 44 are in the United States. That is why the US count of 12,933 live entries says little: 12,498 of them are ports on those 39 addresses, about 320 each. In our test further down, most relays from busy addresses came out in another country. The Hong Kong list follows one such address in detail.
Raw counts also vary wildly from list to list. At 11:03 UTC, three other free lists counted 838, 787 and 24,109 proxies for the United States.
Which countries have free proxies
Count only the ordinary addresses, those with fewer than 20 live entries, and the order changes. Indonesia leads with 1,005 entries, more than twice the United States with 435.
| Country | Live entries on ordinary addresses | Of those, flagged datacenter |
|---|---|---|
| Indonesia | 1,005 | 7.2% |
| United States | 435 | 53.3% |
| China | 190 | 24.2% |
| Philippines | 147 | 1.4% |
| Russia | 145 | 15.9% |
| Colombia | 144 | 9.0% |
| Brazil | 131 | 3.8% |
| India | 122 | 18.9% |
| Mexico | 113 | 9.7% |
| Bangladesh | 104 | 1.0% |
| Germany | 94 | 76.6% |
| France | 76 | 67.1% |
The 34 countries with 20 or more such entries held 3,561 of the 4,003. The other 78 countries shared 442, so most of the world has only a handful of free proxies at any moment. Every country still has its own list, such as the United Kingdom, and our free proxy list shows them all.
Two kinds of free proxy
Our datacenter flag splits the countries in two. In Japan, Singapore, Germany and France, most entries on ordinary addresses were datacenter servers: 85.2, 81.2, 76.6 and 67.1 percent. In Indonesia, the Philippines and Bangladesh, fewer than one in ten carried the flag.
The datacenter kind passed more of our checks. The median entry passed 64.8 percent of its checks in Germany and 69.0 in Japan, against 27.8 in Indonesia and 27.2 in Bangladesh.
Across all ordinary addresses, 21.7 percent carried the flag. With the 44 busy addresses added back, 80.8 percent of the live list did. So the common claim that free proxies are all datacenter servers mostly describes those busy addresses.
Where they exit
Our server in the United States tested all 34 countries twice, at least half an hour apart. Each time it drew up to 50 rows on ordinary addresses and up to 20 on busy ones. Through each row, it asked ip-api.com where the request came from.
On ordinary addresses, about one row in four passed our request on: 339 of 1,360 in the first pass, then 320 of 1,358. More than nine in ten of those relays exited in the listed country, 310 and then 299.
| Country | Relayed, of 100 rows tried | Of those, exited in the country |
|---|---|---|
| Indonesia | 4 | 4 |
| United States | 49 | 45 |
| China | 13 | 12 |
| Philippines | 11 | 11 |
| Russia | 41 | 39 |
| Colombia | 10 | 10 |
| Brazil | 21 | 20 |
| India | 26 | 26 |
| Mexico | 17 | 17 |
| Bangladesh | 31 | 31 |
| Germany | 31 | 19 |
| France | 24 | 17 |
A long list is not a working one. Indonesia had the most entries on ordinary addresses, yet only 4 of its 100 rows relayed. In Russia, 41 did.
Fifty relays from ordinary addresses left their country, and 30 of them came from Germany, France, Ukraine and Finland. Germany alone sent 12 of its 31 relays abroad, 4 of them to the United States.
Busy addresses sent nearly every relay elsewhere: of their 42 relays, 4 exited in the listed country. The busy Hong Kong address gave 27 of them, exiting in Japan, Taiwan, Singapore and the United States. By the second pass, an address in Pakistan had turned busy, and its 9 relays exited in eight other countries. The busy US addresses relayed 4 of the 40 rows we tried.
Check the padlock too. In each pass, 17 of the 50 ordinary US rows failed a certificate check. Nearly all showed an expired certificate for Google or Cloudflare. Our direct check saw valid ones, so something along those paths had replaced them. The United States list has the details.
Picking a proxy for one country
Start from the country's own list and skip rows that share one address with many others. Where few rows relay, as in Indonesia, expect to try dozens before one works. Then check where the proxy really exits before you rely on it. Our proxy checker reports status, speed, anonymity, country and network. Our guide on how to check if a proxy is working shows the one-line test.
Treat the town next to a row as a guess. MaxMind, a geolocation company, estimates 99.8 percent accuracy at country level but 66 percent for US cities within 50 km.
When free is not enough
A free proxy is fine for a quick look at how a page appears from another country. For search results by city, prices you will act on, or anything that must work tomorrow, it is the wrong tool. Our guide to proxies for SEO covers the search case. Our page on free residential proxies explains why free lists rarely hold home connections at scale.
Our residential proxies exit through home connections in the country you pick, and residential traffic starts at $0.44/GB.
How we measured
The counts come from one read-only query of our pool at 10:59 UTC on 28 September 2026. An entry is live when it passed our last check, and an address with 20 or more live entries counts as busy. The datacenter flag is the public field our list API serves. Our free proxy data study covers the pool as a whole.
The exit test ran the same day from 11:00 to 12:43 UTC, from our server in the United States. Besides ip-api.com's echo, it opened Cloudflare's trace page and Google through each row over encrypted connections. That is where the certificate failures showed.
Sources
- MaxMind: geolocation accuracy: country and city accuracy of IP geolocation.
- ip-api.com: JSON API documentation: the service our test asked where each exit sits.
- Cloudflare: the /cdn-cgi/ endpoint: the trace page our test fetched through each proxy.
