Double NAT at Home: Which Public IP Apps Actually See
TL;DR / Summary: In a double NAT setup, two routers each run their own private network, and only the outer one holds the public address. Every device sees that same public IP, which is why lookups agree with each other and still fail to explain why incoming connections cannot find a machine inside.
Two routers, two private ranges, one public address
A typical home with one router has a single private network, usually 192.168.1.0/24, and a single public address assigned by the provider. Add a second router for better Wi-Fi coverage or a separate office network, and you create a second private range behind the first.
Devices on the inner router get addresses from its own pool. Their default gateway is the inner router, which in turn has an address on the outer router’s network. The inner router performs its own translation before handing traffic to the outer one.
That stacking is what the name describes. The public address still belongs to the outer device, so every device on either network leaves through the same public IP.
The chain a packet walks before it leaves
Start with a laptop on the inner network. It sends a request to its gateway, the inner router. The inner router rewrites the source to its own outer-facing address and forwards the packet. The outer router then rewrites again to the public address and sends it upstream.
On the way back, each router consults its translation table and reverses the rewrite. The tables are keyed by port numbers and timers, which is why a long-idle connection can die while a fresh one works.
From the outside, the entire house looks like one machine. A public IP lookup from the laptop and a lookup from a phone on the outer network return the same address, because both ultimately exit through the same device.
What changes when the outer link is a carrier gateway
Sometimes the outer device is not yours. A fibre termination unit, a fixed-wireless receiver, or a provider-supplied gateway may already be doing NAT before your router sees the connection. In that case the public address may not even be on your equipment.
This matters because the outer router’s status page might show a private address such as 100.64.x.x while the public lookup shows something else entirely. That pattern usually means carrier-grade NAT upstream, and it limits what any port forwarding inside the house can achieve.
Check the outer router’s WAN address before blaming the inner configuration. If the WAN address is private, the public address is held further upstream and the chain has three layers, not two.
Port forwarding in a double NAT house
A forwarding rule on the inner router alone does nothing for the internet. Traffic arrives at the outer router first, and if no rule there points it to the inner router, the packet is dropped.
The working pattern is a rule on both devices: the outer router forwards the port to the inner router’s WAN address, and the inner router forwards it to the target device. Give the inner router a reserved address on the outer network, or the rule breaks the next time it reboots.
- Reserve the inner router’s address on the outer DHCP server
- Forward the port on the outer router to that reserved address
- Forward the same port on the inner router to the device
- Retest from outside the network, not from a phone on the same Wi-Fi
Testing from inside is the classic false positive. Many routers handle their own public address in a special way, so an internal test succeeds while an external attempt fails.
Why the lookup result looks wrong but is consistent
People often expect a device behind a second router to report a different public address, then conclude the lookup is broken when both report the same one. The result is correct: the public address is a property of the exit point, not the device.
What differs is the private address, which the lookup never shows. If you need to distinguish devices for logs or access rules, the public IP cannot do it. You need something at the application layer, such as an authenticated session or a unique token.
Geolocation can look wrong for the same reason. The public address belongs to the provider’s pool, which may be registered in a city far from the house, especially on mobile or fixed-wireless links.
Flattening the chain without breaking the network
The cleanest fix is usually to stop routing twice. Put the second device in access-point mode so it bridges Wi-Fi clients onto the outer network, or connect it by cable to a LAN port with DHCP disabled. Then there is one private range and one translation layer.
If the second router must stay a router, keep its network genuinely separate and accept the consequences: forwarding needs two rules, device discovery across the boundary does not work, and anything relying on broadcast traffic will fail.
A public IP check such as My IP Inquire is still useful in both designs, because it confirms the exit point. It just cannot tell you how many routers stand between the device and that exit.
Deciding whether the second router earns its place
Write down what the second router provides: coverage, isolation, a lab network, or a guest boundary. If the answer is coverage, access-point mode gives the same Wi-Fi without the forwarding puzzle. If the answer is isolation, double NAT is doing real work and should stay.
Either way, label the devices and record the reserved addresses. The next time a game console cannot accept connections, that note saves an hour of guessing about which layer is refusing the packet.