You clicked "Check proxy" in the dashboard, it showed the connection works and the region looks right — and the next day your account started throwing captchas. This is the real experience of many sellers after buying a proxy: treating "it connects" as "it passes." Connectivity is only the lowest bar for a proxy, just like a car inspection can't stop at whether the engine starts. Key takeaway: Judging proxy IP quality means passing four gates — speed (latency and packet loss), purity (fraud score and risk value), anonymity (DNS/WebRTC leaks), and stability (long sessions that don't drop or drift). Every gate has free tools to test it. This article walks the four gates in the order of "screen before buying → inspect after buying → recheck before going live," with steps you can follow directly.
A "connectivity test" only answers whether the proxy is reachable; IP quality answers what is behind that proxy. The two are often conflated, and the price is account risk.
The proxy test button in an anti-detect browser (like MasBrowser's Check proxy) verifies three things: whether the proxy server connects, whether the credentials are correct, and whether the exit region is the one you want. It cannot detect these hidden problems:
Platform risk control sees all four problems, while a connectivity test sees none. So buying a proxy can't be a one-question exam — pass the four gates below one by one (if you're new to proxies, the anti-detect browser proxy setup guide covers selection through integration; take that lesson first, then come back to inspect).

The core of a speed test isn't "is it fast" but "is it steady" — average latency is only the entry ticket; packet loss and jitter are what decide the experience.
To test proxy speed, pair two free tools (third-party tool links carry nofollow, same below):
Reference values (based on everyday backend work): below 200ms feels smooth, 200-300ms is acceptable, and above 400ms with packet loss the lag is obvious; an IP with long-term packet loss above 3% will break both page loads and account operations. For batch operations (multiple environments running at once), hold packet loss to an even stricter standard than a single store.
What an IP has been used for in the past decides how platforms treat it now — that's where "dirty IPs" come from, and it's the whole point of purity checks.
Two tools, each with its own job:
Here is a trap almost no tutorial will tell you: the checkup sites are watching you too. ping0.cc was caught in packet captures by the technical community: when you visit the page it uses WebRTC to grab the real IP behind your proxy, and even pairs your IPv4 and IPv6 records. In other words, if you check "is this proxy clean" with a bare browser, your own real network information gets taken instead.
The right way is to test from an isolated environment: create a dedicated checking environment in an anti-detect browser, bind any cheap proxy line to it, and set WebRTC to disabled — then the checkup site receives the environment's information, not your real exit. This is itself a live drill of "guard against the testing tool too."
The standard for anonymity is concrete: does the "you" that platforms see expose anything beyond the proxy. Leak tests are how you find those cracks.
Ranked by how often they appear, the three most common leak points:
This testing is a different thing from fingerprint checking: fingerprint checks look at whether browser parameters are self-consistent; leak tests look at whether the network layer bypasses the proxy. Only when both are clean is the environment truly clean — for fingerprint consistency, see the workflow in verifying environment isolation with ip-api and BrowserScan; this article won't repeat it.
The first three gates are a static checkup; stability is dynamic observation — many proxies ace every static metric and show their true colors after 24 hours online.
Stability testing needs no special tools, only time:
With this, the four gates close the loop. The whole acceptance process strings into five steps: screen out lossy and detouring routes with ping.pe before buying → check purity with Scamalytics/ping0 after buying → check DNS/WebRTC leaks in the environment → run a 24-hour stability test → only after everything passes let the real accounts go live. The first four steps take under an hour in total, yet they avoid roughly ninety percent of network-layer accidents after launch — never skimp on the proxy fee, and never skimp on testing.

Rule of thumb: below 200ms is smooth, 200-300ms is acceptable, above 400ms replace it. But packet loss matters more than the absolute number — never use an IP with long-term loss above 3% no matter how low the latency, and an evening-peak retest exposes problems best.
Community reference points: Scamalytics 0-14 is low risk, above 25 replace it; ping0 risk value 0-25 counts as clean, 70+ will basically be blocked by platforms. Remember a score is a "reference tool," not a "verdict" — for important accounts, cross-check with multiple tools.
That's normal. Each site has different data sources, update frequency, and focus — some look at ASN ownership, some infer from routing traits. When results disagree, go with the majority, and trust the dimension most relevant to your target platform first.
You can test them, but it hardly matters — the problem with free proxies isn't "can you detect they're dirty," it's "they're dirty the moment you detect them": IPs from public lists have been used by countless people and are wiped out at the purity gate. The test workflow is better suited to accepting paid proxies; the risk boundaries of free proxies are analyzed in the risks and right choices of free proxy IPs.
Proxy IP quality acceptance is essentially turning "hope it's fine" into "confirmed it's fine." Speed, purity, anonymity, stability — four gates, all free tools, a one-hour process, in exchange for not having to watch captchas with your heart in your mouth after launch.
If you're about to roll out proxies in bulk, first create a dedicated checking environment in MasBrowser (official site) — WebRTC disabled, fingerprint based on IP — and run this article's test flow with it; it's both safe and handy. The free plan includes 2 environment slots — download MasBrowser, vet the next proxy before it goes on duty.
