We run a verification engine that pulls from 116 free proxy source lists, checks every address it finds, re-checks the survivors continuously, and archives the ones that die. By November 2, 2026, it had made 426,593,121 verification attempts against 954,218 distinct addresses.
That is a large enough sample to answer a question nobody selling proxy lists wants asked: how much of this actually works?
The short version
487,839 of the 917,593 proxies in our dead archive never completed a single successful check. Not once, ever. They were published on a list, we found them, we checked them, and they never answered. That is half of every address we have ever found.
Finding 1: half of all free proxies never worked, not even once
The usual framing is that free proxies "stop working". For half of them, there was never a working state to stop from.
The funnel, 954,218 addresses
| Stage | Addresses | Share |
|---|---|---|
| Discovered across 116 lists | 954,218 | 100% |
| Archived as dead | 917,593 | 96.2% |
| Of those, never once succeeded | 487,839 | 51.1% of all |
| Still in active rotation | 36,625 | 3.8% |
| Answered within the last 24 hours | 28,174 | 3.0% |
| Passing their most recent check | 6,793 | 0.7% |
HProxy verification engine, measured November 2, 2026.
The two bottom rows lead straight to the next finding.
Finding 2: "working" and "working right now" differ by four times
28,174 addresses answered a check at some point in the last 24 hours. 6,793 passed their most recent one. Both numbers are true at the same moment, and the gap between them explains the whole experience of using a free proxy list.
A free proxy flaps. It answers, then it goes quiet, then it answers again, on a cycle nobody controls. A list that reports "30,000 working proxies" is reporting the set that worked today, while the set working this second is about a quarter of that. Pull ten addresses off such a list at random and expect two or three to respond.
This is also why testing matters more than sourcing. Any address you take from any list is a hypothesis until you check it, which is what our free proxy checker is for.
Finding 3: the source you pull from matters enormously
We track which list every address came from. The spread in quality across those 116 sources is far wider than we expected when we started recording it.
Best and worst source lists by live yield
| List | Addresses | Yield |
|---|---|---|
| Best list | 8,569 | 23.9% |
| Second | 2,786 | 22.9% |
| Third | 5,982 | 19.1% |
| Middle list | 126,732 | 3.8% |
| Largest list | 546,164 | 1.19% |
| Low yield | 136,553 | 0.65% |
| Worst list | 315,972 | 0.32% |
HProxy verification engine, measured November 2, 2026.
Read the bottom row again. One source list has published 315,972 addresses over its lifetime. 1,025 of them are alive. That is a hit rate of about one in 300, and it is a list people actively use.
Note the inverse relationship in that table. The sources contributing the most addresses tend to have the worst yield, because volume on a free proxy list comes from scraping other lists, and nobody verifies the result. The ten largest lists in our sample average a 1.5% yield. A list of 200,000 addresses usually holds the working proxies of a list of 1,000, plus a great deal of noise, which is the argument we made in the biggest free proxy lists before we had this data to back it.
Finding 4: the ones that do work last about ten days
Of the proxies that worked at least once and then stopped, half were gone within 235 hours of the day we first saw them, about 10 days. We count until a proxy stopped answering; archiving comes a day or more later. And we count only addresses with fewer than 20 such proxies, leaving out Amazon's networks and one June evening when our checker failed. The figure was measured on 2026-09-28.
We corrected this finding three times, twice on September 27, 2026, and once on September 28, 2026. Its first version said six days, because it also counted proxies that never worked at all. For those, our record only marks the first failed check, so they have no lifespan to measure.
The second version said about 25 days. It counted every port as its own proxy, and a few hosts carry thousands of ports. 804 addresses with 20 or more such proxies held 88 percent of that sample, and one network in Iran held half of it. Counting only the other addresses left 35,888 proxies on 29,141 addresses and a median of about 15 days.
The third version still counted one evening in June. Our checker's test target failed that evening, and about 128,000 entries were recorded dead at once. It also counted entries on Amazon's networks, where no free proxy relayed a request in our September country tests. Without both, 25,977 proxies remain, and that is the figure above.
That number fits what a free proxy usually is. Almost none are deliberate public services. They are misconfigured servers, forgotten test boxes, and open relays that nobody meant to leave open, so the lifespan of a free proxy is really the time it takes for somebody to notice and close it.
A median of about ten days does not make old lists safe. Half of all entries never worked at all, and a list that does not say when each address was last verified gives you no way to tell the working ones from the rest.
Finding 5: where they are, and what they leak
Among addresses alive in the last 24 hours, the United States leads at 12,045, ahead of Indonesia at 3,775, Switzerland at 832, India at 685 and Brazil at 551. The first two hold 56 percent of that set.
If your job needs any other country, the free pool is thinner than the headline count suggests.
Anonymity is the other half of it. Among the addresses passing their latest check, 3,804 graded elite and 1,405 graded transparent. A transparent proxy forwards your real IP in the request headers, so the destination site sees the proxy address and yours. About one in five of the live pool leaks the thing most people are using a proxy to hide. We covered the mechanics of that in are free proxies safe.
What changed since August
We first published this study in August 2026. The same counts on November 2 give the figures below. The share that never worked fell by ten points, and the share of live proxies that leak the user's address rose by seven.
The same counts, three months apart
| Measure | Aug. 11 | Nov. 2 |
|---|---|---|
| Addresses | 589,918 | 954,218 |
| Checks | 129,131,311 | 426,593,121 |
| Never worked | 61.4% | 51.1% |
| Passing now | 4,023 | 6,793 |
| Transparent | 13.7% | 20.7% |
HProxy verification engine, measured August 11 and November 2, 2026.
What this means if you actually want to use free proxies
- 1
Verify every address yourself, always
Half of all published addresses never worked at all. Treat any list as a set of candidates, and check before you use, every time.
- 2
Judge a list by freshness
One list in our sample holds 315,972 addresses and has a 0.32% hit rate. If a list does not publish a last-verified timestamp per address, its size tells you nothing useful.
- 3
Expect about one in four to answer at any moment
The gap between alive-today and alive-now is about four times. Pull more candidates than you need and assume most are asleep.
- 4
Check the anonymity grade before sending anything
About one in five live free proxies is transparent and forwards your real IP. Never put credentials or personal data through an unverified one.
- 5
Use free for testing, paid for anything that matters
Free proxies are genuinely useful for confirming a page renders from another country. For work with a deadline, a 0.7% instantaneous hit rate is a poor foundation.
Method
Every figure here comes from our own verification engine, read on November 2, 2026, unless a line gives another date. The engine pulls from 116 source lists, 103 of them currently enabled, checks each discovered address, re-checks survivors on a rolling schedule, and moves addresses that stop responding into a dead archive. Keeping that archive is what makes the never-worked-once figure computable at all.
Three definitions, because they carry the whole study:
- Discovered means a distinct ip:port pair our engine has ever recorded, whether it ever answered or not.
- Alive within 24 hours means the address completed at least one successful check in the last day.
- Passing its most recent check means the single latest check on that address succeeded. This is the strictest reading and the one that produces 6,793.
Cumulative checks (426,593,121) counts every verification attempt ever made against every address, live and archived. Our retained per-check history table holds a rolling window of them, so that total comes from the per-address counters.
These are measurements of one engine's view of the free proxy world on one day. Another engine pulling different sources would see different numbers, and ours will look different next month. The date is on every figure for that reason.
If you want the practical version of all this, are free proxies safe covers the risks and how to check if a proxy is working covers verification. Our own continuously re-checked free proxy list is the output of the engine described above, and every address on it carries the time it was last confirmed alive.
