Updated: April 2026
IP geolocation technology has been around for a long time. It is the only non-intrusive tool a service provider can use to estimate the geographical location of online visitors, and it has proved itself as the main engine for content localisation, digital rights management, customer targeting, and fraud detection.
But how accurate is an IP geolocation lookup? Can we trust it to make critical business decisions?
Accuracy has been one of the most debated topics in this space for years. One of the most popular myths is that IP geolocation is inaccurate because it is based on public data and usually points to an organisation's headquarters rather than the real user's location. The reality varies widely — not all IP geolocation services are built the same, and they use different technologies to source their data. See our article IP Geolocation Demystified for more details on how conventional services operate.
This article addresses the fundamentals of IP geolocation and explains the best accuracy we can theoretically expect, the conceptual limitations to be aware of, and to what extent we can trust the data we receive.
IPv4 is here to stay
When we refer to IP geolocation, we are primarily considering the IPv4 address space, because most internet traffic still comes from IPv4 addresses. IPv6 was created to solve the global shortage of IPv4 address space, but it has not happened even though two decades have passed since it was introduced. The main reason is probably the most surprising one — there isn't a real shortage of IPv4 addresses after all.
To some extent there is, as it is almost impossible to get IPv4 addresses for free anymore. But how can roughly 3.1 billion publicly routable IPv4 addresses service our entire internet, while there are more than 5.5 billion internet users, 4.5 billion smartphone users, between 19 and 21 billion IoT devices, and millions of public IP addresses used by global network infrastructure? Visit our IPv4 Address Space Monitoring for up-to-date figures.
The answer is that we only need a dedicated, publicly routable IP address when we require two-way communication — when we need to be able to reach out and also receive incoming connection requests from the outside world. Most of our internet activity is one-directional. When we browse the internet, we initiate the connection ourselves and don't expect websites to stay actively connected to our device.
There are plenty of technologies to share an IP address for one-way communication: Network Address Translation (NAT) at home or in small offices, Carrier-Grade NAT (CGNAT) for mobile operators, and name-based virtual hosting for websites. We tend to classify IP addresses as static or dynamically allocated, but we often don't get an address assigned to us exclusively — we only temporarily make use of it, and possibly share it with other users simultaneously.
The old and widespread assumption that every single IP address has a single device behind it and can therefore be tracked to an exact geographical location is not valid. When we consider IP geolocation, we must first consider how an IP address is actually being used.
Static IP addresses
A static IP address's geolocation is the easiest to get right. Regardless of the exact method, IP geolocation is always evidence-based — it is always about field evidence: real-world observations of which IP address was seen from which location. Not all evidence is correct; it could be way off, just like a faulty GPS reading. The quality of any IP geolocation service depends on what evidence it has access to and how it handles conflicting signals.
When we have fresh, consistent, and precise location evidence for a static IP used by a stationary device, the resulting accuracy can be very high — often within a few metres in raw data. This is why BigDataCloud deliberately rounds coordinates to the nearest kilometre, capping the maximum reported precision to within approximately 1 km². IP geolocation is not suitable for identifying a person's exact location, and it should not be.
Dynamic IP addresses
A dynamic IP is one your ISP assigns temporarily. It may stay the same for months if you rarely reconnect, or change every few hours depending on the provider's policies. The longer the same IP address remains at the same physical location, the better the chance that a geolocation service can accumulate enough consistent evidence to locate it as accurately as a static one.
Dynamic IPs are typically drawn from a pool managed by the router serving your section of the network. That router has a pool of IP addresses it allocates to customers — sometimes sequential blocks, sometimes several separate ranges. The router serving you directly is essentially responsible for your section of the network, and that section usually corresponds to a distinct geographic boundary — a service area.
The maximum geolocation error you should expect when using an IP address that was previously witnessed somewhere else is the maximum distance from your physical location to the outermost boundary of that service area. This is what BigDataCloud calls the confidence area — a critically important piece of information that tells you where else the IP address could be if it was assigned dynamically. For decisions that depend on location accuracy, the confidence area must be considered alongside the point estimate.
Cellular networks
IP address allocation by a cellular operator can loosely resemble a fixed-network ISP's dynamic allocation, but there are at least three differences that make IP geolocation significantly more challenging for mobile networks.
First, the device behind the IP address is not stationary — a user can drive several hundred kilometres and still be using the same carrier-assigned public IP address. Second, users often don't have exclusive use of an IP address; they are sharing it with other customers, sometimes very remote ones. Third, cellular operators can consolidate several internal network sections behind a single shared pool of IP addresses, making the effective service area considerably larger than a fixed ISP's pool.
Even a theoretically perfect IP geolocation service cannot always estimate the correct, up-to-the-moment location for every IP address on a cellular network. Even the cellular operators themselves often don't have this data in real time without consulting internal logs.
What a good geolocation service should do — and what BigDataCloud does — is identify whether an IP belongs to a cellular network, and provide a confidence area that honestly reflects the scale of uncertainty. The point estimate indicates the most probable location based on accumulated observations, which naturally skews toward larger population centres. But the confidence area may span an entire state or country, and for informed decision-making, both are needed.
Hosting networks
Another important categorisation is distinguishing between IP addresses that are serving a human-operated device directly, and those serving a server, bot, or middlebox. At BigDataCloud we categorise these as consumer and hosting networks.
Hosting is the overarching term for all sorts of unattended computers — those running from data centres, as well as office or private property servers. Hosting network IP addresses are those that power public websites, mail servers, VPN services, Tor, proxy networks, and many other applications, both legitimate and malicious. Residential proxy or VPN networks, which route traffic through consumer IP address blocks, are also classified as hosting.
Most hosting IP addresses are static allocations and describe the location of the infrastructure itself rather than any specific end user.
Edge cases
Roaming cellular
When you travel overseas and turn roaming on, websites may still recognise you as being in your home country — even if you are thousands of miles away. Most mobile operators tunnel their roaming customers' traffic back to the home network, primarily to control billing of data services. Technically this is very similar to a VPN routing your traffic home.
This makes IP geolocation particularly challenging for roaming users. On top of regular CGNAT — where multiple scattered mobile customers share the same IP address — roaming adds the possibility that some of those users are located anywhere in the world. BigDataCloud can sometimes detect signals of overseas usage, but since the same IP is simultaneously being used by home-country customers, we report the home country location. Statistically, that is the correct answer for that IP pool.
VPNs, proxies, and Tor
Users who mask their IP address through a VPN, proxy, or Tor exit node will be geolocated to the location of the exit node rather than their actual location. In these scenarios, accurate classification of the IP as an anonymiser is often more useful than the coordinates themselves.
Conclusion
IP geolocation is categorically not suitable for tracking the exact geographical location of a person or device. However, in the majority of cases, it gives meaningful and reliable insight into the geographic area where an IP address is being used. The confidence area, network type classification, and confidence level together make IP geolocation-based decisions considerably more robust than a point estimate alone.
| IP type | Characteristics | Typical accuracy |
|---|---|---|
| Static | Permanently assigned; consistent evidence over time | Very high; privacy-capped to ~1 km² |
| Dynamic | Temporarily assigned from an ISP pool; may be reassigned | Moderate to high; use confidence area to understand the range |
| Cellular / CGNAT | Shared by many users across a wide area; device is mobile | Point skews to population centres; confidence area may span a state or country |
| Hosting / infrastructure | Servers, VPNs, proxies, Tor exits | Describes infrastructure location, not a specific end user |
For both a point estimate and a clear picture of the uncertainty around it, use the IP Address Geolocation with Confidence Area API. For a detailed explanation of confidence levels, read How to interpret Confidence Value. For current, independently verifiable accuracy figures across network types and regions, see the Daily IP Geolocation Accuracy Report.