Why Does Your IP Show the Wrong City? How Geolocation Databases Treat VPN Servers

This week I connected to a VPN server in Estonia and looked up the exit IP. Four of the five geolocation sources I checked put it in Tallinn. The fifth put it in Stockholm, in another country, about 380 km away. Nothing about the server had changed between lookups. Only the database was different.

That gap is the reason an IP address "shows the wrong city", whether it belongs to a VPN or to your home router. This article explains where the gap comes from, then shows what a small test of five VPN servers turned up.

Where geolocation databases get their locations

An IP address is a number assigned to a network. Nothing in it says "Frankfurt". Every location you see in an IP lookup is a database's best guess, built from a few kinds of input.

Registry records.

Regional internet registries record which organization holds each block of addresses, and that record includes an address and a country. It is a reasonable starting point for country and a poor one for city, because an ISP's registered address is often its headquarters, not where the customers are.

Self-published feeds.

Network operators can publish their own mapping of IP ranges to locations. The format, called a geofeed, is described in RFC 8805, and RFC 9632 explains how to attach a feed to a registry record. RFC 8805 notes that Google already uses these feeds in its geolocation pipeline. They are accurate when the operator is honest and careful, and nobody checks them by default.

Measurement.

Some providers send pings and traceroutes to an address from many places and work out where it must be from the round-trip times. IPinfo describes this approach for its own data in its VPN location report. It is the hardest signal to fake, because latency is limited by the speed of light.

Behavioral signals.

Some vendors add data from apps and devices that report a location while using a given IP. This helps most on residential broadband and least on shared or mobile networks.

Every provider blends these inputs differently. That is why two lookups of the same address can disagree, and why none of them can tell you which one is right.

Why your own IP can show the wrong city

You do not need a VPN to get a wrong city. A few ordinary things do it:

• Your ISP's pool is regional. Addresses are often assigned from a range registered to a city or region other than yours.
• Mobile networks move addresses around. MaxMind's accuracy overview says a mobile IP can serve phones hundreds of kilometers apart within minutes, so providers return a broad region instead of a city.
• Records go stale. IPinfo says about 30% of region-level IP data changes every year. A block that moved last month can still show its old city.
• Databases fall back to a default. When a provider knows only the country, it returns a fixed point for that country. For years, MaxMind's default point for the United States sat on a farm in Kansas, and more than 600 million IP addresses pointed at it. The people living there were visited by law enforcement and accused of fraud they had nothing to do with. MaxMind later moved its default points into bodies of water.

The Kansas case is the extreme version of something that happens quietly all the time: a precise-looking coordinate that actually means "somewhere in this country".

The test: five VPN servers, five lookup sources

I wanted to see how much geolocation sources disagree on addresses that are known to be VPN servers.

Method. On 23 October 2026, I connected through a commercial VPN service to servers labeled Estonia, Netherlands, USA, Germany, and Singapore. For each connection I recorded the exit IP and looked it up on IPLocation.net, then cross-checked it against ipinfo.io, ip-api.com, ipwho.is and the free DB-IP API. I recorded the country, city, network owner and the VPN or proxy flag that IPLocation.net shows.

Limits. This is five addresses on one day, with one IP per server. It shows how sources behave. It does not measure how good any VPN or database is. I had no independent measurement of where each server physically sits, so "agreement" below means sources agreeing with each other and with the label on the server.

The last octet of each IP is masked. ipwho.is matched to the IPLocation.net city in all five cases, so it is left out of the table to keep it readable.

What the results show

Country was nearly unanimous. Of 25 lookups, one got the country wrong: ip-api.com put the Estonian server in Stockholm. It is the same kind of disagreement MaxMind describes when it says country accuracy is high but never perfect.

City disagreements were small and local. Halfweg is a village just west of Amsterdam, and Spokane Valley and Liberty Lake are neighboring towns in eastern Washington. These are the differences you get when two databases place one network on different sides of a metro area. Overall, 21 of 25 city results matched.

The network owner gave the server away every time. None of the five addresses belongs to a home ISP. All five sit on hosting or data center networks, and IPLocation.net labels them "data center". A website that cares about VPN traffic can read that as easily as the city.

VPN detection was inconsistent. IPLocation.net flagged three of the five addresses: two as public proxies and one as a VPN. It flagged the other two as nothing at all. Both the Estonian and the US servers looked like ordinary hosting to it. If a site decides whether to trust you from a flag like this, the same VPN can pass on one server and fail on the next.

When the label itself is the problem

My five servers were mostly where their labels said, as far as the sources agree. That is not guaranteed.

Providers can list locations in countries where they have no hardware. The server sits in one place, and the operator tells registries and geofeeds that it is somewhere else. Databases that trust self-published data then repeat the claim. RFC 9632 even says that when geofeed data are present, other location hints in the registry record should be ignored, which is what makes a false declaration stick.

IPinfo tested this directly. It connected to 20 popular VPNs and collected more than 150,000 exit IPs across 137 countries. By its measurements, 17 of the 20 providers had traffic exiting in a different country than advertised. In the 736 cases where its measured country disagreed with other widely used datasets, the median distance between them was about 3,100 km. The study also found 38 countries that were claimed by at least one VPN but never appeared as a real exit country for any of them.

This is IPinfo's own research and its own measurement platform, so treat the numbers as one vendor's view. The direction is still hard to dismiss. A city, or even a country, in a VPN app is a label until something measures it.

What a VPN changes, and what it doesn't

A VPN replaces your IP with the server's. The website then sees the server's location and network owner. MaxMind says as much: with a VPN, it can locate the server, not the person using it. If you want a refresher on what an IP address reveals and who can see it, the basics are there.

Three things a VPN does not fix:

• A wrong city on your home IP. It swaps one address for another, and the new one can be placed just as imprecisely. My Estonian server was "Stockholm" in one database.
• Being recognizable as a VPN. As the test shows, hosting-network ranges are easy to spot and detection varies by vendor.
• Tracking that ignores your IP. IPLocation.net's article on what your IP address reveals covers browser fingerprinting, which survives a change of address. For a plain-language view of what a VPN does and does not hide, start there.

For disclosure: I write for a VPN company. That is why I tested on a neutral question: how databases treat these servers, and why this article does not tell you which service to use.

How to check your own IP in five minutes

1. Open the IP lookup with VPN off and note the country, city and network owner.
2. Connect to a server and run the lookup again. Check that the network owner changed, not just the city.
3. Compare two or three sources. If one disagrees on country, treat that source as the odd one out.
4. Connect to a second server in the same country and repeat. If the city changes between servers, the site you are using may be reading a different range.
5. Run the lookup from a phone on mobile data. You will usually get a broader region and a different city than you do on Wi-Fi.

If a site shows you the wrong city after a VPN connects, try another server in the same country before you assume anything is broken. The cause is nearly always the database, not the connection.