You look up your IP address and the result puts you in a city two hours away. Or a different site says you are in another country entirely. Nothing about your connection changed; the sites simply asked different databases, and the databases gave different answers.
The same thing happens to proxy buyers, with money attached. If you pay for residential proxies targeted by country, the question of where an exit address “is” decides whether you got what you paid for. This guide explains where IP locations come from, why they disagree, and how Atlas checks them.
There is no official location for an IP address
Regional internet registries record which organisation an address block is allocated to, along with that organisation’s registered address. That is a billing and contact record. It says nothing about where the device using the address actually sits.
Everything beyond that record is inference. Every “IP location” you have ever seen is a commercial or community database’s best estimate, built from indirect evidence and refreshed on its own schedule.
Where geolocation databases get their answers
Each database blends several sources and weighs them differently:
- Registry and WHOIS records. The starting point, and often the country of the company rather than the user.
- Routing data. Which network announces a block, and from where.
- Geofeeds. Files some internet providers publish that map their own address ranges to locations. Accurate when kept current, often stale when not.
- Latency measurements. Probes in known locations time their round trips to an address and narrow down where it could be.
- Device and app signals. Location reported by apps and sites with the user’s consent, tied to the address seen at the time.
- Hostnames. Reverse DNS names that contain city or region codes.
- Corrections. Reports from users and network operators that an entry is wrong.
Two databases with different access to these sources, different weights and different refresh cycles will reach different conclusions about the same address. Neither has to be careless for that to happen.
Why the answers disagree
Addresses move. Home internet addresses are usually assigned dynamically, and providers regularly move blocks between regions as they grow. A block that served one city last month may serve another now. Each database catches up at a different time, and until it does, it reports the old location.
City is much harder than country. A country rarely changes for a block; a city does. When a database lacks good evidence, it often falls back to the provider’s main hub, the regional capital, or a default point near the middle of the country. That is how a user ends up “located” in a city they have never visited.
Many homes share one exit. Some providers route a whole region’s subscribers through shared gateways. The database sees the gateway’s city, which can be far from the household.
Relays hide the user. Corporate VPNs, privacy relays and satellite links put the exit address somewhere other than the person. Databases that flag these handle them differently from databases that do not.
Errors persist unevenly. A wrong entry fixed in one database stays wrong in another until its next refresh, which can be weeks away.
Why your own IP shows the wrong location
If your own address is placed in the wrong city or country, the practical steps are limited but real:
- Check several lookups, not one. If most agree and one is off, that one database is the problem.
- Submit a correction to the database that is wrong. Most vendors accept them.
- Ask your internet provider whether it publishes a geofeed. Providers that do can fix their own entries at the source.
- Wait a refresh cycle. Corrections reach each database, and then each website using it, on different timelines.
Fixing one database does not fix every site, because each site buys its own.
Why it matters more for residential proxies
When you request a country through a proxy gateway, three different views of location are in play: the provider’s routing claim, what the geolocation databases say, and what the target site decides. Only the last one changes your results.
If a target places your exit in the wrong country, you get the wrong prices, the wrong search results, the wrong language, or a block page for traffic that looks out of place. The response may still come back with a normal status code, so the failure is easy to miss. A pool sold by country is only worth what it delivers in that country, which is why accurate geo-targeting belongs in any provider evaluation alongside success rate and speed.
How Atlas measures geolocation accuracy
Atlas records the exit address the target saw on every request, then looks up each unique address in an IP database and counts the share placed in the country the requests asked for. Addresses the database has no country for are left out rather than counted as misses. The figure is per unique address, not per request, so one heavily reused address cannot dominate the result.
That share is the Authenticity part of the Atlas Score, set against the best network tested. You can see the Authenticity sub-score for any two networks on the side-by-side proxy comparison, and the exact rule in the Atlas definition of geolocation accuracy.
Two limits are worth stating:
- Country only. City is where databases disagree most, and a single database’s city call is weak evidence either way.
- One database. A mismatch can be the database’s error rather than the provider’s. Across thousands of addresses, isolated errors average out; a consistent pattern does not.
Doing it yourself
You can run the same check on any provider in an afternoon:
- Request one country through the gateway and send requests to an endpoint that echoes back the address it saw. Record what the target saw, not what a dashboard reports.
- Build the set of unique exit addresses.
- Look up each address in two or three databases. A free lookup such as Shifter’s IP Info geolocation API is one option.
- Sort each address into three groups: every database agrees on your country, every database agrees on another country, or they disagree.
- If you target cities, repeat at city level and expect far more disagreement.
- Finally, test against the sites you actually care about and check that they serve local content.
For a broader test plan that covers location alongside speed and success, see how to test residential proxy speed, success rate and location accuracy.
What to make of a mismatch
A few addresses placed in the wrong country is normal, and usually the database’s fault. Treat it as noise.
A consistent pattern is different. If a large share of exits sold as one country agree, across databases, on a neighbouring country, the addresses are probably not where the provider says. That is a sourcing or routing question to put to the provider, with your list of addresses attached.
If the databases disagree among themselves on most of your exits, the pool may be fine and the databases simply behind. In that case the target site’s own behaviour is the only verdict that counts.