Carrier-Grade NAT Signs on Residential IP Lookups

Carrier-Grade NAT Signs on Residential IP Lookups

Abstract global network visualization over a cityscape

TL;DR: If inbound connections fail, open ports on your router do nothing, and your public IP is shared with strangers, you may be behind carrier-grade NAT (CGNAT). Confirm the WAN address your modem shows versus what an IP inquire page reports, then plan allowlists and hosting accordingly.

What CGNAT changes about “my public IP”

Classic home setups put a public IPv4 on the ISP modem/router WAN. Port forwards reach your LAN. CGNAT inserts another translation layer: many customers share one public IPv4. Your modem’s “WAN IP” might be a private or shared CGNAT pool address (often 100.64.0.0/10). The address websites see is the carrier’s outer IP—not uniquely yours.

Symptoms that lookup pages alone cannot explain

  • Port forward tests time out despite correct LAN rules
  • Game consoles complain about Strict NAT after every modem swap
  • Self-hosted VPN endpoints never accept inbound peers
  • Allowlisting “your” IP blocks neighbors when the carrier remaps

A public IP inquire still shows an address—it just is not a personal inbound door.

Compare three numbers

  1. Address on the modem WAN status page
  2. Address shown by a browser IP lookup
  3. Address reported by a router “stun” or WAN detect widget, if any

If (1) is 100.64.x.x / 10.x / 172.16–31.x while (2) is a public ISP IP, CGNAT (or at least double NAT) is in play. If (1) and (2) match a public IP yet inbound still fails, look at ISP blocks or double NAT from an extra router instead.

Why allowlists misbehave on CGNAT

Outbound SaaS traffic from you and from other subscribers may share egress. You are not “hacking” when a neighbor’s traffic originates from the same public IP—your carrier designed it that way. Conversely, your egress IP may change more often as the carrier rebalances pools, breaking brittle allowlists.

Prefer application logins, mutual TLS, or IPv6 unique addresses over raw IPv4 allowlists when CGNAT is confirmed.

Distinguishing CGNAT from double NAT in one pass

Both break inbound connections, and both show a mismatch between what the modem reports and what a lookup page returns. The difference is where the second translator lives. Double NAT comes from your own equipment—an ISP modem in router mode feeding a second router, each running its own private subnet.

Test by connecting one machine directly to the modem’s LAN port and repeating the three-number comparison. If the mismatch vanishes, your own second router is the culprit and the fix is local, usually bridge mode or a single routing layer. If the mismatch persists on the bare modem, the translation sits upstream and belongs to the carrier.

That single test saves an unnecessary support call and, more importantly, prevents buying hardware to solve a problem the hardware cannot touch.

Record which layer the mismatch came from in the same note as the three numbers. A support conversation that begins “the translation is upstream of my equipment, here is the test that shows it” moves far faster than one that begins with a description of the symptom.

Ask the technician to confirm the finding in the ticket even when they agree verbally. The written line is what makes the next support contact thirty seconds instead of thirty minutes.

Latency and MTU symptoms worth noting

CGNAT adds a translation hop, which shows up as slightly higher and occasionally spikier latency under load, especially during peak hours when the shared pool is busy. A path MTU discovery failure can also appear as connections that establish but stall on the first large transfer.

Neither symptom proves CGNAT on its own—congested Wi-Fi produces similar readings—but together with a private WAN address they form a coherent picture. Note them in the same log entry so the evidence accumulates in one place instead of getting retested from scratch each time.

IPv6 as an escape hatch

Many carriers offer native IPv6 with globally routable prefixes even when IPv4 is CGNAT’d. If your homepage or SSH can bind to v6 and the peer supports it, inbound becomes feasible again. Verify that your inquire tool and the remote peer both see the same v6 address family you configured.

Hosting and lab implications

Home labs behind CGNAT need relay solutions: reverse SSH, cloud-hosted entry nodes, or vendor relay agents. Do not keep buying “gaming” routers expecting magic public ports the carrier will not map to you.

Document CGNAT in your runbook so future port-forward guides get skipped faster.

Talking to the ISP without theatrics

Ask whether your line uses CGNAT and whether a static public IPv4 or dual-stack option exists (sometimes paid). Provide the WAN vs public lookup mismatch. Skip accusing them of “hiding your IP”—bring the three-number comparison.

Recording CGNAT evidence with www.myipinquire.com

Screenshot the inquire result beside the modem WAN page on the same day. Retest after reboot; note whether the public IP stayed sticky. Keeping a short log via www.myipinquire.com makes CGNAT diagnosis boring and repeatable—exactly what you want before redesigning remote access.

Similar Posts