Geolocation Data Explained: Accuracy, Signals, and Real World Use

IP geolocation sounds simple on paper: map an IP address to a place on Earth.

It’s anything but.

Between leased IP space, routing changes, anycast, VPN “virtual locations,” and constantly shifting infrastructure, the internet doesn’t behave like a tidy database. And that’s exactly why building accurate geolocation data requires more than just looking up a record, it requires continuous verification. 

In a recent IPXO webinar, I walked through how geolocation datasets are built, what makes them difficult to maintain, and how we use measurement to validate the signals we collect. Watch the full video for more details.

The Simple Goal of Geolocation

At its core, IP geolocation is the process of mapping an IP address to a geographic location.

That location might be high-level (country or region), or more granular (city, and in some cases, something closer to facility-level). But geolocation doesn’t come from one perfect source. It’s built by collecting many imperfect hints.

A hint can be useful. A hint can be wrong. A hint can be outdated. And sometimes, a hint can be intentionally misleading. The job of a geolocation data provider is to take all of these signals and figure out which ones are consistent with reality.

There are multiple ways to derive hints about where an IP address might be located. In the webinar, I walk through several categories of hints, including:

• Geofeeds, a standardized format that allows network operators to publish IP prefix-to-location mappings
• WHOIS and registry metadata, which sometimes includes useful country-level context
• Reverse DNS, where naming conventions can occasionally reveal patterns (including airport or city codes)
• Network paths, where traceroute-like observations can provide additional context

However, hints alone are not enough. You still need to figure out which hints are trustworthy, and whether they even make sense together.

Why Validation Matters

This is where measurement becomes essential. At IPinfo, we validate geolocation claims using latency measurements, which allow us to cross-check location hints against what the internet is actually doing.

A simple way to think about it:

• If a location claim is accurate, latency patterns should generally align with that geography.
• If it’s not accurate, you can sometimes prove it’s impossible.

That’s because packets are still constrained by physics. The speed of light (and the reality of fiber) gives us a hard upper bound on how fast signals can travel.

In the video, I share an example where a “country” claim doesn’t hold up when you measure it. It’s one of those moments where you realize why verification is so important, especially when the signals look legitimate on paper.

How ProbeNet Validates

To do measurement-based validation at scale, you need global points of presence. That’s why we built ProbeNet, IPinfo’s internet measurement platform, which allows us to continuously run active verification and collect evidence across the internet. As of January 2026, ProbeNet consists of:

• 1,300+ points of presence
• Coverage across 140+ countries and 540+ cities
• A footprint designed to include common regions and harder-to-reach locations

We use ProbeNet not just to improve geolocation, but also to support broader verification across our data pipelines. And one advantage of having our own measurement network is flexibility: we can deploy new measurement approaches as research evolves.

Anycast: One IP, Many Locations, One Headache

One of the most common “edge cases” in geolocation is anycast, where the same IP address can be announced from multiple locations, and traffic is routed to the nearest instance.

If you don’t account for anycast, it can look like:

• geolocation inconsistency
• random location changes
• “drift” over time
• disagreement between providers

In the webinar, I explain how large-scale measurements can be used not only to detect anycast, but also to begin enumerating where anycast instances exist. 

VPN Virtual Locations

Another concept we discussed in the webinar is VPN location claims, specifically the difference between:

• a VPN that actually has infrastructure in a given country
• and a VPN that claims a country presence via a “virtual location”

In our recent VPN study, we analyzed a large set of VPN exit behavior across 20 providers, and found widespread evidence of traffic exiting in a different country than advertised.

That’s not just a problem of inaccuracy. For users, they’re being misled about the country they’re selecting. What’s actually happening in the network is altogether different. And the compliance issues are fraught. 

Geolocation Is a Moving Target

If I had to boil geolocation down to one challenge, it’s that the internet is dynamic, and geolocation is a moving target.

To stay accurate, you need more than static metadata. You need an evidence-based approach that can keep up with constant change, from shifting routing and leased IP space, to anycast deployments and VPN location claims that don’t hold up under measurement.

And while we didn’t go deep on them in this blog post, the webinar also covers three important themes that sit just beneath the surface of “accuracy”:

• Freshness: how quickly geolocation must update to reflect real-world changes
• IPv6: why measurement strategy has to evolve as IPv6 traffic continues to grow
• DNS signals: how reverse DNS can provide useful context, but can also introduce noise when it’s stale or misleading

If you want the full walkthrough of how geolocation hints get collected, how we validate them, what breaks accuracy assumptions, and why measurement changes everything, watch the full IPXO webinar.