PublicIPChecker

Guide

How IP geolocation works

6 min read · Updated 2026-09-28

The four data sources behind every 'your IP is in Berlin' result, realistic accuracy numbers by network type, and the reasons your city often shows up wrong.

How the databases are built

No geolocation service has a pin on your house. They combine four imperfect sources:

  1. Registry allocations — the internet's regional registries (ARIN, RIPE NCC, APNIC, LACNIC, AFRINIC, APNIC) assign address blocks to ISPs, and those registrations carry organisation addresses, typically at city level. This is the foundation every database starts from.
  2. Network inference — BGP routing data and observed traffic patterns let providers estimate where a block actually serves, which often disagrees with where the organisation is registered.
  3. Active probing — test clients in datacenters around the world ping candidate networks and use round-trip time to triangulate where a block physically reaches.
  4. User-contributed data and models — opt-in samples ("your device says you are in city X, and your IP is in this block") plus machine learning that fuses everything.

Every commercial database (the ones this site and its peers use, and the big named ones) mixes these in different proportions — which is why two services can disagree on the city.

How accurate to expect

GranularityTypical accuracy, fixed broadband
Country~95–99%
Region / state~80–90%
City~70–85%

On mobile, satellite, CGNAT and corporate networks, city accuracy drops noticeably — often to "the carrier's gateway city" rather than yours. VPN and proxy traffic is, by definition, wrong: it shows the exit server's location. Treat the checker's location as strong evidence, which is why it is labelled an estimate.

Why your city is often wrong

  • CGNAT — your traffic exits the carrier at its gateway city, which may be a hundred kilometres from you (what is CGNAT).
  • Server registration vs reality — the block is registered to the ISP's head office city, not your cabinet.
  • Mobile and roaming — the tower serving you may be a neighbouring operator's, in the next city.
  • VPN / proxy — the result is the server's city, full stop.
  • Stale data — blocks move between operators; databases refresh at different rates.

What geolocation must not be used for

IP location is an inference, not a position fix. It cannot find a street, a building or a person, and it degrades exactly where it is most likely to be misused — border areas, mobile connections, shared addresses. Legitimate uses are regional (localised pricing and language, compliance with regional law, fraud priors); illegitimate ones treat a ~80% city guess as a physical location. If a service claims to know where you are "exactly" from your IP alone, you are being lied to.

Check yours: the checker shows the country, region and city alongside the ISP and ASN — if the city surprises you, the ASN usually explains why (carrier gateway, datacenter exit, or your ISP's registration city).

Frequently asked questions

Can IP geolocation pinpoint my house?

No. The best realistic resolution on fixed broadband is city-level, and usually it is the carrier's or the datacenter's city, not yours. GPS or an address is needed for physical location.

Why does a different service show a different city for my IP?

Each database blends registry data, probing and inference differently, and refreshes on different schedules. Country-level results usually agree; city-level ones often do not.

Is my city data 'personal data'?

Combined with other data, an IP address (and its location) is treated as personal data under GDPR and similar laws — which is why this site does not log your results server-side.


Keep reading

Check any other IP address

Investigating a suspicious login, a spam email header or a server log entry? Run any IPv4 or IPv6 address through the same geolocation and proxy checks.