When you open almost any website, a database lookup occurs in the background that converts your IP address into a set of geographic coordinates. The result powers streaming geo-restrictions, fraud detection, regional pricing, compliance rules, and targeted advertising. But how accurate is that lookup — and what can it actually reveal about you?
The short version: country-level accuracy is excellent. City-level accuracy is mediocre. Street-level accuracy is nonexistent. And several common situations routinely produce completely wrong results, regardless of how good the database is.
How IP Geolocation Databases Are Built
IP addresses are allocated hierarchically. IANA (the Internet Assigned Numbers Authority) distributes large address blocks to five regional registries (ARIN, RIPE NCC, APNIC, LACNIC, AFRINIC). Those registries allocate smaller blocks to ISPs and organisations, which in turn assign individual addresses to subscribers.
The registration records — called WHOIS data — contain the country and often the city of the organisation that holds the block. Geolocation database vendors, such as MaxMind, IP2Location, and ipinfo.io, harvest these records continuously. But raw WHOIS data is only the starting point.
Vendors supplement registry data with several additional signals:
BGP routing table analysis. The Border Gateway Protocol route announcements that keep the internet interconnected contain information about where traffic for a given prefix is actually being routed. This helps pinpoint cases where a block is registered in one country but actively routed through another.
Active probing and traceroute data. Some vendors run network probes that measure latency from known vantage points. If pings from a New York server reach an IP in roughly 5ms and pings from London take 100ms, the IP is probably not in London despite what the registry says.
User-submitted corrections and crowd-sourced data. Mobile apps and browser scripts that users have consented to can report GPS coordinates alongside IP addresses. Over millions of samples, this builds a rich signal about which IP ranges appear in which real-world locations.
Commercial data partnerships. ISPs and mobile carriers sometimes license location data directly to database vendors under commercial agreements — particularly for mobile IP address pools.
Despite these methods, the databases are imperfect snapshots of a dynamic network. IP allocations change. ISPs reassign blocks. Companies merge. The records lag reality.
Accuracy by Connection Type
The connection type dramatically affects how accurate geolocation can be.
Fixed residential broadband tends to produce the best results. The address is assigned to a fixed location, the ISP's WHOIS registration usually reflects the correct region, and the address rarely moves. Country accuracy approaches 99%. City accuracy is roughly 70–80% for dense urban areas, lower for rural addresses.
Mobile networks are the most problematic. Carriers use large pools of addresses shared across entire regions, routing all mobile traffic through a handful of gateway servers that may be hundreds of kilometres from the subscriber. Geolocation may report the city where the gateway is located, not where the phone is. In practice, mobile IP geolocation is reliable at country and state level, but city-level accuracy can be as low as 30–40%.
Business and enterprise connections often route through a corporate headquarters or regional data centre regardless of where the employee is sitting. An employee working from Edinburgh may appear to geolocate to London if their traffic exits through a London data centre.
VPN and proxy users are deliberately masking their location. The geolocation database correctly identifies the VPN exit node, not the user's physical location. This is precisely what VPN users want — and exactly why IP geolocation cannot be relied upon as a definitive identifier of physical presence.
CGNAT (Carrier-Grade NAT) assigns the same public IP address to hundreds or thousands of subscribers simultaneously. Any location derived from a CGNAT address represents the location of the carrier's NAT gateway, not any individual subscriber.
Why Your Detected Location Might Be Wrong
Beyond the structural limitations above, several specific situations routinely produce wrong results:
Legacy IP block allocations. IP blocks allocated decades ago to organisations at one address are sometimes transferred or reassigned without the registry records being updated promptly. A block registered to a New York firm may have been sold to a Berlin company and is now in active use in Germany, but the WHOIS entry still shows New York.
Satellite internet. Providers like Starlink route traffic through ground stations. Your IP may geolocate to the ground station's location, which could be in a different country from you. Early Starlink deployments caused widespread confusion as users in Europe geolocated to US cities.
Split tunnelling and proxy services. Corporate proxies, content delivery networks, and even basic ISP traffic optimisers can cause your traffic to exit at an unexpected point. The IP seen by the destination server belongs to the proxy or CDN node, not your actual access point.
IPv6. IPv6 allocations are newer, and geolocation databases have less historical data to draw on. IPv6 geolocation accuracy is generally lower than IPv4, particularly at city level, though this is improving.
What IP Geolocation Cannot Tell Anyone
Understanding the hard limits is as important as understanding what works.
It cannot identify you as an individual. An IP address, even accurately geolocated, maps to a network endpoint — not a person. Multiple people in a household share the same public IP. CGNAT means thousands of people share one IP. Even with a precise location, there is no mapping from IP to identity without additional data.
It cannot determine your physical location within a building or street. The practical floor is city-district level, and only under ideal conditions. Street address, floor, or room are completely beyond what any IP-based system can determine.
It cannot overcome deliberate obfuscation. A residential VPN service assigns your traffic a residential IP in another city or country. The geolocation database will correctly identify that city as the source — because that is the correct IP-level fact. It simply is not your real location.
It cannot determine intent or identity. A journalist using a VPN, a child using a parent's device, and a corporate employee working remotely may all have IP addresses that geolocate somewhere entirely different from their physical and legal location.
This matters because many systems use IP geolocation as a proxy for certainty it does not provide: age verification, location-based access controls, fraud detection. These systems catch the majority of cases but have an irreducible error rate rooted in the technical realities above.
FAQ
Can a website know my exact street address from my IP?
No. IP geolocation does not and cannot provide a street address. The best it can achieve is an approximate city or district, and only under favourable conditions (fixed residential broadband, accurate database entries, no proxying). Any website claiming to know your street address from your IP alone is either using additional browser-based signals (like the Geolocation API, which requires your permission) or is simply wrong.
Why does my IP show a city 50 miles from where I actually am?
The most common reasons are: your ISP routes traffic through a regional hub in a different city, your CGNAT gateway is located elsewhere, or the geolocation database has a stale record. This is normal and not a sign of a misconfiguration. It simply reflects that IP geolocation is an approximation based on network routing, not a GPS signal from your device.
Does using a VPN completely hide my location from websites?
A VPN replaces your real IP with the VPN server's IP, so websites will geolocate you to the server's location rather than yours. This is effective for defeating IP-based geolocation. However, websites can still determine your rough location through browser-level signals: the Geolocation API (if you have previously granted permission), Wi-Fi-based location (if the site uses it), and timezone settings visible in JavaScript. A VPN handles IP geolocation; it does not neutralise all location signals.