IP geolocation
The inference of physical location from an IP address, reliable at country level and frequently wrong at city level, and the mechanism proxy geo-targeting relies on.
IP geolocation is an educated guess dressed as a fact. There is no location stamped inside an address; what exists is inference, built by combining the registration records that say which organisation holds a range, observations of how traffic to that range is routed, and data the networks themselves sometimes publish. Commercial providers assemble those signals into databases mapping ranges to places, and every site that greets you with your city is reading one of them.
The accuracy has a sharp gradient that governs all practical use of it. Country-level is genuinely dependable, above the high nineties for the major databases, because address blocks are allocated to networks operating in known countries and that fact is hard to get wrong. City-level is a different animal: often it reflects where the provider registered the range or where its routing infrastructure sits rather than where the subscriber physically is, so it can be a different city, sometimes hundreds of kilometres off, while looking perfectly precise on screen.
That gradient is exactly why proxy targeting is priced and structured the way it is. Country targeting is the reliable product because the underlying inference is reliable; city targeting is harder to supply and harder to verify because the databases the destination consults may not agree that your exit is in the city you selected. You can hold a genuine residential address in Lyon and still be read as Paris by a site whose database is stale for that range, which is a database disagreement, not a proxy fault.
The staleness point deserves weight because it drives a whole class of confusing results. Databases are refreshed on the vendor's schedule, ranges get reassigned and relocated between refreshes, and different sites license different vendors updated at different times. So the same exit can geolocate to one place on one site and another place on the next, and neither is lying, they are reading different maps of different ages. Recently reassigned ranges are the usual culprit behind an address that tests correct in one checker and wrong in another.
For anyone verifying a proxy this yields a concrete method rather than a single reading. Check the exit against several geolocation sources, not one, because one source is one vendor's current opinion; agreement across several is far stronger evidence than a confident answer from any single database. And treat a country match as solid and a city match as probabilistic, which is simply reading the tool at the accuracy it actually has.
How HProxy handles it
Our IP lookup reports location as what it is, an inference with a confidence that falls off below country level, rather than a single authoritative city. That honesty is also why our targeting leans on country as the dependable unit and treats city as best-effort, since promising city precision the underlying databases cannot confirm would be selling a number we cannot stand behind.
Frequently asked questions
How accurate is IP geolocation?
Very accurate at country level, above the high nineties for the major databases, and much weaker at city level. City data often reflects where a provider registered the range or sited its routing rather than where the user is, so it can be off by a different city or hundreds of kilometres while appearing precise. Trust country as fact and city as a probable guess.
Why does my proxy show a different city than I selected?
Usually a database disagreement rather than a proxy fault. The site consulted a geolocation vendor whose data for that range is stale or was never precise below country level, so it places your genuine exit elsewhere. Recently reassigned ranges are the common cause, which is why the same address can read correctly in one checker and wrongly in another.
How do websites know my location from my IP?
They look the address up in a commercial geolocation database, which was built by combining registration records, routing observations and network-published data into a range-to-place mapping. Nothing about location is stored in the address itself; the site is reading a third party's inference, and the quality of that inference is exactly what varies between vendors and over time.
How do I verify a proxy's real location?
Check the exit, not the gateway, against several geolocation sources rather than one. One source is a single vendor's current opinion; agreement across several is much stronger evidence. Treat a country match as reliable and a city match as probabilistic, and expect some disagreement on recently reassigned ranges, which is normal rather than a sign the proxy is wrong.
Back to the full glossary.