Proxy Detection Test

What platforms can and cannot see

A website learns about a visitor from three places: the connection, the request and the page.

• The connection. Your IP address and everything that can be looked up from it: the network that announces it (see the IP type checker), an approximate location, and whether it appears on reputation lists or known VPN, hosting or Tor ranges. The TLS handshake adds a client fingerprint, which the JA4 fingerprint checker shows.
• The request. Headers such as User-Agent, Accept-Language and client hints arrive with every request, before any script runs.
• The page. Scripts read the time zone, languages, screen, canvas, WebGL, fonts and automation flags (the browser fingerprint test lists them), can gather WebRTC candidates, and can record how you type, scroll and move the pointer.

Through a proxy, a website cannot see your SIM, your phone number or a device behind a carrier's NAT, and it cannot see your own IP address unless something leaks it: WebRTC, DNS or an app that bypasses the proxy. Mobile apps are a separate case, since they can read device identifiers that a web page cannot, and this test does not cover them.

Account history weighs heavily as well. A new account, a device the platform has never seen and a first login from an unfamiliar network add up, whatever a checklist says.

How each check is decided

• Network type. Inconsistent when the announcing ASN is on our hosting, cloud and CDN list, or its registered name says hosting. Carrier and ISP networks pass. Few real visitors browse from hosting ranges, so many sites challenge them.
• WebRTC. Public candidate addresses from the browser are compared with the address of the HTTP request. A different address on another network is inconsistent. A different address inside the same ASN passes, because carrier NAT can give different flows different public addresses. The method follows the IETF's WebRTC IP handling requirements.
• Time zone. The browser's offset now against the offset of the zone our hosting edge assigns to your IP.
• Languages. Whether the browser's languages include one commonly used in the IP country. This is the weakest check, because English is common everywhere.
• Accept-Language. The header our server received against navigator.languages. A difference usually means only one of them was overridden.
• Automation. navigator.webdriver, a headless token in the User-Agent, User-Agent against client hints, and a phone User-Agent without touch support.

A signal that cannot be read on a run is shown as not assessed and left out of the count, so the summary never treats a missing measurement as a pass.

Why there is no single trust score

Each platform weighs signals its own way, retrains its models and adds signals no outside page can see, such as account age, payment history and behaviour. A score computed here would be our weighting presented as theirs. A number also hides the reason: it does not tell you whether to fix the time zone or the network. The checklist shows each observation with its fix and leaves the weighting where it belongs.

Passing every check is not a guarantee. It means none of the common contradictions showed up on this run. Can mobile proxies be detected and how websites detect proxies go through what platforms do beyond these checks, and why a proxy gets detected covers the usual causes.

Fixing the common failures

• Network type shows hosting. You are testing a VPN, a datacenter proxy or the proxy's entry server rather than its exit. Route the browser through the proxy and run the test again from inside that session. A dedicated mobile proxy exits from a real carrier line with its own SIM, used by one customer only.
• WebRTC exposes another address. Set the browser or antidetect profile to replace or disable WebRTC, so it cannot reveal an address outside the proxy. The WebRTC leak test shows every candidate in detail.
• Time zone or language mismatch. Match them to the proxy location, as described in fixing a mismatch.
• Automation flags. navigator.webdriver is set by WebDriver-based tools by design, as the WebDriver standard requires, and headless modes add their own markers. A site that sees the flag is being told the session is automated, so check the site's terms before automating it.

After a fix, run the test again in the same browser session. A result from your everyday browser says nothing about an antidetect profile or a remote browser.