What is my IP address, and why every checker tells you a different one

An IP address is a number assigned to a device on a network. For the public Internet, two versions exist: IPv4 (32 bits, written as something like 203.0.113.42) and IPv6 (128 bits, written as something like 2001:db8::1). Most of what end users call 'my IP' is the public IPv4 address that their Internet provider routes traffic through.

The public IPv4 you see is not your computer's address. Your computer sits behind a router that does network address translation (NAT). The router has one public IP from the ISP, and every device in your house shares it. When a website sees your request, it sees the router's IP — not yours. The router then forwards the response to the right device using local-network addressing.

If you're on a corporate or VPN network, the public IP is one further hop away: the VPN's exit node, or the corporate proxy's gateway. From the website's perspective, the request came from that exit — not from your home. This is the source of the disagreement between different IP check tools: they see different network paths to the same destination.

Single-IP echo services — ifconfig.me, ipinfo.io, ip.sb, ipapi.com — all do roughly the same thing: they read the source address of your HTTP request and show it to you. If you're connecting without a proxy, they all show your home IP. If you're behind a proxy, they show the proxy's exit — but only if the proxy is the actual TCP-layer egress.

The disagreement appears when your network has more than one IP. A common case: a VPN whose client routes only some traffic through the tunnel. ifconfig.me goes through the tunnel (because the proxy rule 'foreign sites' includes it) and shows the VPN exit. ipapi.com is a US-hosted service also covered by the rule and shows the same exit. ip.sb is on a different region rule and shows your home IP.

This is not a bug in any of the tools. It's the tools honestly reporting what each one sees from the path it sits on. To figure out the real picture, you need a checker that hits multiple destinations and shows you each one — that's what IPCombo does with its connectivity matrix.

Beyond the obvious IPv4, two more addresses commonly hide behind your connection. IPv6 is the modern replacement — most ISPs give you one but websites still mostly serve over IPv4. If your network is dual-stack, an IPv6-only endpoint will see your IPv6, which is usually different from your IPv4 and may not be the one your proxy exits from.

CGNAT (Carrier-Grade NAT) is the second twist. Many ISPs put entire neighborhoods behind a single public IPv4. From the website's perspective, hundreds of households share the same IPv4. This makes IP-based geolocation imprecise at best, and is one of the reasons 'is this IP mine' is harder to answer than it sounds.

If you want to know which IPv4 you actually have, hit a service that bypasses your proxy rules (or run a test from your browser's incognito with the proxy off). IPCombo shows both your IPv4 and, when present, your IPv6 alongside the WebRTC srflx address, so you can spot CGNAT (multiple WebRTC IPs in the same /24) and the IPv4/IPv6 split in one screen.

An IP can be looked up against a GeoIP database to give an approximate country, region, city, ISP, and ASN. The country is usually within 99% accuracy at the country level. Region is maybe 80% accurate. City is closer to 50-70% — most IP-to-city databases are wrong about a third of the time. ISP and ASN are essentially exact.

An IP cannot tell you who is using it. An IP belongs to whoever the ISP assigns it to, but the ISP doesn't know who's behind it. The website knows which IP made the request, but not which person. Even a residential IP doesn't mean a real human — proxies rent residential IPs from ISPs and resell them. This is why 'is this IP residential' cannot be answered from the IP alone.

What the IP can tell you, when combined with the ASN and latency: the class of the network. A Linode or AWS IP in the 10000-30000 ASN range is datacenter. An IP in a Comcast or AT&T range is residential. An IP in T-Mobile or Verizon mobile ranges is mobile. The latency pattern is corroborating: datacenter responds in single-digit ms, residential in 20-80ms, mobile in 50-200ms. The ASN is the source of truth — the latency is a sanity check.

When IPCombo finishes a test, you see one or more public IPs at the top — one per exit your network actually has. If you see a single IP, your network has a single public exit. If you see two or three, you have multiple — a VPN tunnel + your home IP, or a dual-stack setup, or a CGNAT split that the browser's WebRTC probes were able to surface.

Below the IP, you'll see the geo: country, region, city, and ASN. Country and ASN are reliable. Region is approximately right. City is best-effort. For an AI or streaming service that cares about region matching, country + ASN is what matters — city rarely affects routing decisions.

The most useful follow-up is the connectivity matrix. For each target in the matrix, IPCombo tells you which exit your request actually came from. If 'google.com' shows your home IP and 'openai.com' shows your US exit, you've confirmed that your proxy is whitelisting Google but proxying OpenAI — exactly the evidence no single-IP tool can give you.