Arguments about free proxies are usually anecdotal, so we measured ours. We run a free proxy list as a tool on this site: a community-sourced list, refreshed every 30 minutes, exactly like every other free list on the internet. On 12 September 2026 we took the 100 proxies it was serving at that moment and tried every one of them from a server of ours.
The method
The list gave 100 unique addresses across 97 different network blocks, so this is not one broken host repeated. Each proxy got two requests with an 8 second timeout: one to an HTTPS endpoint that echoes back the address it sees, and one to the same endpoint over plain HTTP. A proxy counts as working only if a request completed and returned a valid address.
The results
| Outcome | Count |
|---|---|
| Completed an HTTPS request | 2 |
| Plain HTTP only, HTTPS failed | 10 |
| Failed both | 88 |
| Median response time of the 12 that worked | 4.6 seconds |
| Slowest working proxy | 8.0 seconds |
| Exited from a different address than the one listed | 5 of 12 |
Two numbers matter more than the headline. The first is that only 2 of 100 could carry an HTTPS request, which rules out almost every real task, because almost everything worth fetching is served over HTTPS. The second is 4.6 seconds at the median. A paid residential request to the same endpoint returns in well under a second, so even the survivors are roughly an order of magnitude slower.
Why free lists decay so quickly
Most entries on a free list were never intended to be proxies. They are misconfigured servers, forgotten test boxes and consumer devices with an open port, found by scanning the internet. Nobody maintains them, and they vanish when an administrator notices or a router reboots. That is why every free list refreshes constantly: it is a snapshot of what answered a scan minutes ago, not a pool with an operator behind it.
It also explains the 5 of 12 that answered from an address different to the one advertised. Traffic through an unknown relay is going somewhere you cannot see, which is a poor foundation for anything you care about.
What free proxies actually cost
Think in terms of the job, not the price tag. Say you want 10,000 text-only pages.
- With the free list. 12 usable proxies, most HTTP-only, median 4.6 seconds per request, dying as you go and requiring a re-scrape of the list every half hour. You will write retry logic, a health checker and a rotation layer before you collect a single useful row, and the run will still fail overnight.
- With paid residential. 10,000 text-only pages is roughly 1 GB with images and scripts blocked, which is about $1.75 on Budget Residential. The addresses answer on the first attempt and do not need a health checker.
The free option is not cheaper. It costs an afternoon of engineering to avoid spending less than the price of a coffee, and it fails in ways that are hard to debug because the infrastructure changes under you. The same arithmetic in more detail is in cost per successful request: the right metric is never dollars per gigabyte, it is dollars per request that actually worked.
When a free proxy is the right tool
There are real cases, and they share one trait: nothing breaks if the proxy dies mid-request.
- Checking whether a site serves different content, prices or language from another country.
- Confirming your code's proxy configuration is correct before you pay for anything.
- Learning how proxies work, including what a slow or dead one does to your error handling.
- A single manual fetch of a public page where an address has no consequences.
If you use one, treat it as hostile infrastructure. Send nothing unencrypted, never send credentials or cookies, and assume the operator logs what passes through. Our guide on using a free proxy list safely covers the rules, and the free proxy list guide explains where these lists come from.
What to look at instead of the price
When comparing paid providers, the numbers worth asking for are the ones our test exposes: what share of requests complete, how long they take, whether the address you were promised is the address you exit from, and whether the pool is the same an hour later. Price per gigabyte only matters once those are settled. On our own products that means $1.75 per GB for budget residential, $2.75 for the premium pool, and $7.90 per 30 days for a static address that stays yours. Whichever provider you pick, run the test we ran: a few hundred requests to the site you actually care about, counting successes rather than reading marketing pages.
Frequently asked questions
Do free proxies actually work?
A small minority do, briefly. We tested 100 proxies from a public list on 12 September 2026: 88 failed outright, 10 handled plain HTTP but not HTTPS, and 2 completed an HTTPS request. The 12 that worked answered in 4.6 seconds at the median, against well under a second for a paid residential IP.
Why do free proxy lists go stale so fast?
Because most entries are not services at all. They are misconfigured servers and devices that were never meant to relay traffic, so they disappear when someone notices or reboots them. That is why every free list, ours included, refreshes every 30 minutes or so. The list is a snapshot of what answered a scan a few minutes ago, not a pool you can build on.
Is it safe to use a free proxy?
For anything unencrypted, no. An HTTP proxy operator can read and alter traffic that is not protected by TLS, and you cannot know who runs a random open proxy. Never send credentials, cookies or personal data through one. In our test, 5 of the 12 working proxies exited from a different address than the one listed, which means the traffic passed through infrastructure we could not identify.
When is a free proxy good enough?
One-off checks where nothing matters if it fails: confirming a site shows different content from another country, testing that your code's proxy plumbing is wired correctly, or learning how proxies work. Any job with a deadline, a login, or more than a few hundred requests needs paid addresses.
How much does it cost to replace a free list with real proxies?
Less than the time spent nursing the free one. Budget Residential is $1.75 per GB, and a gigabyte covers roughly 10,000 text-only pages when images and scripts are blocked. That makes a 10,000 page job about $1.75 with addresses that answer on the first try.