IPv4 and IPv6 in proxies: what changes for sites, apps and detection

IPv4 is the original version of IP addresses: 32 bits, about 4.3 billion combinations, now exhausted. That's why carriers, especially mobile ones, have many customers share the same IPv4 through CGNAT. IPv6 uses 128 bits: there are so many addresses that each connection gets a whole block (prefix), usually a /64.

Adoption is growing but varies a lot by country: in April 2026 about half of Google's users connected over IPv6, with peaks around 73% in France and figures around 17% in Italy.

A dual-stack connection has both addresses. When you open a site, the system asks DNS for both the IPv4 and the IPv6 address: if the site has an IPv6 it's tried first and, if it doesn't answer within a fraction of a second, the connection falls back to IPv4. This mechanism is called Happy Eyeballs and modern browsers use it.

With a proxy, your computer no longer makes this choice: the proxy does. It resolves the domain and opens the connection to the site. So the site sees the proxy's IPv6 if both support it, otherwise its IPv4. Whether your home network has IPv6 doesn't matter.

• Mobile IPv4: behind CGNAT the same address is used by many real users, so platforms are careful before blocking it.
• IPv6 and prefixes: anti-abuse systems reason per block, not per single address. Changing only the last part of an IPv6 doesn't make you look like a different user.
• Cheap “IPv6 proxies” are almost always datacenter blocks: millions of addresses from the same provider, easy to recognise by their ASN. The comparison between proxy types is in Mobile, residential and datacenter proxies.
• A mobile carrier's IPv6 has the carrier's ASN and is the same kind of address an ordinary smartphone gets, so to the site it's just another mobile connection.

A leak happens when part of your traffic doesn't go through the proxy and the site sees your real address. With IPv6 it happens more often than you'd think:

• Apps that ignore the proxy: the browser uses the proxy, but another app or a system service goes out directly, often over IPv6.
• WebRTC in the browser: it can reveal local and public addresses, IPv6 included.
• VPNs or tools that only handle IPv4: IPv4 traffic is covered, while IPv6 leaves your network in the clear.

To check, open What is my IP through the proxy: the IPv4 and IPv6 addresses shown must be the proxy's, not your home ones. The full check, including WebRTC and DNS, is in How to test a proxy.

1. Let the site choose: don't force IPv4 or IPv6 unless you have a specific reason. The proxy uses IPv6 where it helps and IPv4 everywhere.
2. Use domain names, not IP addresses, in scripts and tools: that way the proxy can pick the right version.
3. With SOCKS5, let the proxy resolve domains (socks5h in cURL and many libraries). The differences between protocols are in HTTP or SOCKS5.
4. Avoid outdated tools: SOCKS4 and some old programs don't support IPv6.
5. Check your exit with the proxy checker, which shows the IP and the version used.