"I'm in Shanghai, so why does it say Hangzhou?" "This IP shows the US, but my provider says Singapore." Almost everyone who has used an IP lookup tool has run into a location that looks wrong.
The short version: an IP address carries no location information of its own. Every "location" is inferred. Inference has error, and the error follows patterns. Once you understand them, you can tell whether a "wrong" result matters and whether it can be corrected.
How an IP location is produced
IP geolocation databases (MaxMind, IP2Location and IPinfo are well known) combine several sources. The exact weighting and algorithms are not published, but these are the main publicly known inputs:
- Registration data. IP space is allocated by five regional registries (ARIN, RIPE NCC, APNIC, LACNIC, AFRINIC), and the records list an organization, country and address. That tells you who holds the block, not where the user is.
- Network measurement. Probes in many locations measure latency and routes to an IP and estimate position from network distance.
- Operator self-reporting. Network operators can publish a geofeed file under RFC 8805, declaring which region each prefix serves, and database vendors can use it.
- Other data fusion. Some vendors say they blend in additional sources and apply manual corrections, with details undisclosed.
The "city" you see is a best guess weighted from these clues.
Six common reasons the result is off
1. The IP is registered at headquarters but the user is elsewhere. Carriers often run one national network. A block may be registered in a provincial capital or head office city while its users are spread across a whole region.
2. Mobile traffic exits from a few central gateways. Phone traffic typically aggregates at a regional gateway before reaching the internet, so city-level error is especially large on mobile networks. For the shared-exit background, see CGNAT and Shared IPs.
3. Stale data. After a block is transferred, leased or re-planned, the database may go a long time without updating its location.
4. VPNs, proxies and cloud services. You see the location of the exit server's data center, not the user. That is expected behavior, not inaccuracy.
5. Default coordinates when only the country is known. When data supports only a country-level answer, a database assigns a representative coordinate. There have been public reports of a household in the US that became such a default point and was associated with huge numbers of unrelated IP lookups, after which vendors changed how they handle defaults. It shows that a very specific-looking location is not necessarily a specific one.
6. Databases disagree with each other. The same IP often maps to different cities in different databases. ipscoper's data comes from third-party public APIs (a primary and a fallback source), so a change of source can shift results slightly for the same IP. See IP Checker Tools Compared.
How to read the precision
- Country level: usually fairly reliable.
- City level: varies a lot with region, carrier and database.
- Street or house level: not something IP lookup can do, and not something it should be expected to do.
We don't quote a specific accuracy percentage, because it differs too much across countries and network types, and any single figure would mislead.
When you should care
- The country is wrong. If your target market is the US and the IP shows another country, that is a real problem.
- City-level drift. Usually limited impact. That is a rule of thumb, and it depends on your use case.
- Mismatch with timezone and language. The IP's location alone may not matter. What matters is whether it agrees with the rest of the account's environment; see Timezone, Language and IP Mismatch.
How to self-check and cross-verify
- Run the ipscoper IP lookup and note country, city, carrier and ASN.
- Look up the same IP on one or two other sites and compare.
- Prioritize country and ASN/carrier in the comparison, since they are usually more stable than city.
- Use the environment check to see whether the browser timezone matches the IP region.
If several sources agree on country and ASN and only differ on city, there is usually little to worry about. For what an ASN tells you, see What Is an ASN.
Can it be corrected
That depends on whose IP it is:
- The IP belongs to your own network or a leased line. Ask your network provider to update the registration data and publish a geofeed per RFC 8805. Also file correction requests with the major database vendors, which usually require evidence and take an uncertain amount of time.
- The IP is leased from a provider. Ask the provider to handle it. Requests from individual end users are often not accepted.
- It's just a visitor's IP being misjudged. It can't be corrected. Accept the error and don't treat city-level results as certain facts in your business logic.
Also note that fixing one database does not mean the one a given platform uses will change. Which data a platform relies on is not publicly stated.
ipscoper's limits
We don't do geolocation ourselves. We display results returned by third-party APIs and label them as "approximate area, not a precise physical address." The risk score is also a heuristic estimate; see the methodology page.
Summary
A wrong-looking location is not a bug. It follows from how location is inferred. Treat it as a reference with an error range: trust the country, watch the trend at city level, and don't trust anything at street level. When something critical doesn't match, cross-verify across sources first, then decide whether to pursue a correction. For a fuller guide to reading IP data, continue with What Can an IP Lookup Tell You.