Proxy fraud rarely looks like fraud. It looks like a clean landing page, a big number of countries, an unlimited plan and a Telegram handle nobody answers. Every one of those claims can be tested for a couple of dollars before you risk a real deposit — here is how.
Every other check in this article is secondary to this one. A proxy seller who is confident in the product will let you push real traffic through it for a trivial amount of money, or for none at all, before you deposit anything you would miss. A seller who insists on a $50, $100 or $500 minimum before your first request has arranged things so that you find out what you bought only after the money is unrecoverable.
So the first question is never "how much per gigabyte" but "what is the smallest amount I can spend to run my own target through this?" Look for one of three things: a free allowance, a trial, or a minimum top-up small enough that losing it is an annoyance rather than a loss. If none of the three exist, you are being asked to pay for a promise.
And test with your own target, not with a speed-test page. A proxy that loads example.com
proves nothing. Send the requests you actually intend to send, against the site you actually care
about, and read the response bodies rather than the status codes.
curl -x http://USERNAME:PASSWORD@HOST:PORT \
-s -o /dev/null -w "%{http_code} %{time_total}s %{size_download}B\n" \
https://your-real-target.example/page
Run that thirty or forty times. You are looking for consistency, not a single success. Fraudulent and half-broken services very often work perfectly for the first few requests and then collapse into timeouts, because the pool behind them is much smaller than advertised.
There are only a few honest units in this business: per gigabyte of traffic, per address per month, or per concurrent port. All three are countable. You can measure the bytes you sent, count the addresses you were issued, and see how many connections you were allowed to open. If the invoice and your own measurement disagree, you have evidence.
"Unlimited" is not a unit. Residential bandwidth costs the seller money every single gigabyte it passes, so nobody can sell an unlimited amount of it at a flat price and stay solvent. The cost is always recovered somewhere you were not told about: a thread cap that makes throughput useless, silent throttling after some undisclosed threshold, or an account that is closed for "abuse" at exactly the point where you started using what you had paid for. Unlimited plans are not automatically a scam, but they are always a plan whose real limits are hidden, and hidden limits are the mechanism every proxy scam depends on.
This one is free to check and it eliminates a remarkable number of sellers. Find the support channel — Telegram, a ticket form, an email address, whatever they publish — and ask a specific technical question before you spend anything. Not "is this good" but something that requires knowledge to answer: which protocols the product supports, how long a sticky session holds an address, what happens to unused traffic at the end of the month.
You learn two things from the reply. First, whether anyone is there at all; a support link that goes nowhere is the single most reliable indicator that nobody will be there when your traffic stops working either. Second, whether the answer is precise. Real operators answer narrow questions narrowly. A vague, enthusiastic, keyword-heavy reply to a technical question usually means you are talking to someone reselling a panel they have never looked inside.
A specific warning about protocols, because it is where overclaiming is easiest to catch: if a seller says a product supports everything — HTTP, HTTPS, SOCKS5, UDP — ask them to prove the SOCKS5 endpoint with a working command. Plenty of residential products, including the residential pool sold on this site, are HTTP and HTTPS only. A seller who will not narrow that down is guessing.
After the test, go and look at the usage figure in the panel and compare it with what you measured yourself. The two numbers will not match exactly — proxy accounting includes TLS handshakes, request headers and retries that your own counter may not — but they should be in the same neighbourhood and they should move when you send traffic.
# measure your own side while you test
curl -x http://USERNAME:PASSWORD@HOST:PORT \
-s -o /dev/null -w "%{size_download}\n" \
https://your-real-target.example/page | awk '{s+=$1} END {print s" bytes"}'
Three things should worry you. A usage counter that does not update for hours, so you cannot tell what you are being charged for until the balance is gone. Usage that runs at several times your measured volume with no explanation of what the overhead is. And a panel with no per-day or per-request breakdown at all, only a shrinking balance. Billing you cannot audit is billing you have to take on faith, and the whole point of the exercise is to stop doing that.
Country and city counts are the least verifiable claim on any proxy site, which is why they are inflated so freely. The number is also irrelevant to you: what matters is whether the three or four locations your work depends on actually answer. So request them by name and verify the exit yourself.
curl -x http://USERNAME-country-jp:PASSWORD@gw.pyproxy.com:1111 https://httpbin.org/ip curl -x http://USERNAME-country-de:PASSWORD@gw.pyproxy.com:1111 https://httpbin.org/ip
Take the address that comes back and check it against two independent geolocation databases rather than the seller's own checker page, because geolocation records are frequently stale and a datacentre range can be registered to a country it has never been in. Then do the real test: load a page that only behaves correctly from inside that country and see whether it behaves. If a location is sold but silently falls back to a neighbouring country, this is where you find out, and it takes minutes.
Ask any seller where the addresses come from. The honest answers are narrow: we operate our own network of devices under an SDK with a consent flow, we lease ranges from an ISP, or we resell an upstream pool. All three are legitimate businesses. What is not legitimate is a seller who cannot answer at all, or who claims to own a global residential network of their own while running a rebranded copy of a panel you can find on four other domains at four different prices.
To be straight about our own position, since a page about spotting dishonesty cannot practise it: PyProxy is a reseller. The residential, ISP, mobile and datacenter pools sold here come from upstream providers, and the value we add is the shop around them — one balance, per-gigabyte billing, traffic that does not expire and a support channel that replies. We are not going to tell you we built a network we did not build. The right way to treat that claim, and every claim on this page, is the same: do not believe it, test it.
That is what the free gigabyte is for. Link Telegram to a new account and one gigabyte lands on the balance with no card and no deposit, which is enough to run your genuine target, compare the dashboard figure with your own measurement, check the countries you need and ask support a hard question — all before any money changes hands. Apply the same procedure to every seller you are considering, including this one. The ones who object to being measured are the ones the article is about.
Spend the smallest amount the seller allows, or claim whatever free allowance exists, and run your own target through it before you deposit anything meaningful. A seller who will not let you send a single request until a large minimum has cleared has told you what you need to know.
Usually. Residential traffic costs the seller money for every gigabyte that passes, so an unlimited plan is paid for somewhere else: thread caps, throttling, or accounts closed once you use what you were promised. Pricing in a unit you can count, per gigabyte or per address, is the part you can verify.
Not on its own. Most shops resell upstream pools, PyProxy included. What matters is whether they admit it and can tell you what sits behind the gateway. A seller who claims to own a network and cannot describe it is the one to walk away from.
Ignore the number on the marketing page and request the specific countries you need. Check each exit address against two independent geolocation databases, then load a page that only behaves correctly from inside that country. A location count you cannot reproduce is not a feature.