How we work
Every fact on ProxyRange comes from a public page, a public DNS record or a plan we bought ourselves. Each one shows the date we checked it and links to where it came from.
Prices
A price is the lowest total a new customer pays in one purchase for at least 1, 10 or 100 GB of standard rotating residential traffic, from public self-serve options: pay as you go, or the first month of a plan. Taxes are excluded; mandatory fees are included.
Promo codes are never applied; a price that needs one says so. A crossed-out “was” price with no code or deadline is an anchor, so we show the price you’d actually pay. A one-time offer for new customers, or a price shown only after login, keeps its label and is never called the cheapest.
Where a provider’s website terms allow automated reading, a program reads its pricing page every week. Where they don’t, we read it by hand every month.
Terms
We quote terms word for word, with the clause number, the date we read them and a link. “Not stated” means the provider doesn’t say; we never fill a gap with a guess.
Every week we compare each terms page with the text we kept from the week before. A change is recorded word for word, with an AI summary that publishes only if every quote in it matches the change exactly. Any change to a page we quote testing terms from pauses testing that provider until a person has read the page again.
Where the IPs come from
Residential proxies route traffic through other people’s devices and connections. The sourcing check says what we could establish about where a provider’s IPs come from and who it shares them with, using these methods:
- Infrastructure map: we resolve each provider’s documented gateways in public DNS and follow where they point.
- Documentation fingerprints: identical login formats, ports or documentation across brands.
- Web and company links: the company behind each brand, shared site tags and shared DNS hosts.
- Supply and public reports: the provider’s own statements, its bandwidth-sharing apps and SDKs, and reports by security researchers and authorities that name it.
- Terms and price watch: what changes in its terms, and claims that contradict them.
What Clear, Unclear and Concerning mean
- Clear: we looked with the methods above, the provider says where its IPs come from, no source ties it to abusive supply or says it sells another network’s pool, and its gateway doesn’t point into another brand’s.
- Unclear: the provider doesn’t say where its IPs come from, a source ties it to abusive supply or to another network’s pool, its gateway points into another brand’s, or the evidence doesn’t settle it either way.
- Concerning: at least one high-confidence item, either an authoritative public report naming the provider or a strong technical link to a brand named in one, plus a second, independent item from another source. A person approves every Concerning check before it appears; until then it shows as Unclear.
Shared infrastructure is shown on each page but doesn’t decide the check by itself: many brands sell access through the same networks, and that says nothing new about where the network’s IPs come from.
We describe what we observed, when and how, never what we think a company intends. Every item carries its date, its source and a confidence label: high when it is direct and reproducible, medium when several signals agree, low for a single indirect signal.
Shared infrastructure
A link between two brands needs evidence we can show: the same gateway addresses outside big CDNs, the same DNS target, the same site tag, nameservers on one of the two brands’ own domains, identical documentation or the same company. Weaker hints, such as sharing a large hosting network, are labelled as such.
A person checks the evidence for every link before it appears. A link we haven’t seen for 35 days drops off, because infrastructure changes.
Measured tests
We test a provider only where its terms allow it, or after it says yes in writing, and only on plans we buy ourselves. Free credits never count towards a score.
Each monthly run sends requests through the proxy to our own endpoint and counts what it sees: how many IPs are residential, whether the location matches what we asked for, how many distinct IPs there are, and how long a sticky IP holds. We publish totals only; exit IP addresses are counted, then discarded.
The proxy checker
The checker runs one check at a time and stores nothing: no proxy logins and no IP logs. Its data sources are credited on the checker page.
Dates and sources
Every value shows the date we checked it and links to its source. A value we haven’t been able to check again within 45 days goes back to be read by hand.