Your phone tells you where you are to within a few meters, indoors, in a city, while you are walking. It does this by listening to timing signals from a constellation of satellites twenty thousand kilometers overhead, and it does it so reliably that nobody thinks about it.
Underneath that, continuously, since the 1970s, is general relativity. Not as an interesting footnote. As an operational requirement, applied in hardware, every second, or the system does not work.
GPS Is A Clock, Not A Map
The first thing to understand is that satellite navigation is not really about position. It is about time.
Each satellite broadcasts, over and over, a message saying where it is and precisely when it sent the message. Your receiver picks up several of these and measures how much later each one arrived. Because the signals travel at the speed of light, the difference in arrival times gives the difference in distances, and four such measurements pin down where you must be standing.
The conversion factor is brutal. Light covers about thirty centimeters in a nanosecond. A timing error of one microsecond is a position error of roughly three hundred meters. The satellites therefore carry atomic clocks, and the whole system is a distributed exercise in keeping those clocks honest.
Which is where the physics stops being optional.
Two Effects, Pulling Opposite Ways
A GPS satellite is moving fast — about 14,000 km/h — and special relativity says a moving clock runs slow relative to a stationary observer. That costs the satellite clock roughly 7 microseconds a day.
The satellite is also 20,200 km up, in a measurably weaker gravitational field, and general relativity says a clock higher in a gravitational well runs fast. That gains the satellite clock roughly 45 microseconds a day.
They do not cancel. Gravity wins comfortably. The net effect is that an atomic clock in GPS orbit gains about 38 microseconds per day against an identical clock at sea level.
The engineering response is disarming in its bluntness. The clocks are built to tick slightly slow on the ground, at a deliberately offset frequency, so that once they are in orbit and running fast they keep correct time. The correction is soldered in before launch.
The Part Almost Everyone Gets Slightly Wrong
You will read, constantly, that without this correction GPS would be wrong by about 10 kilometers a day.
The number comes from a straightforward multiplication: 38 microseconds of drift, times the speed of light, is roughly 11 kilometers of distance. It is a good way to convey scale, and it is not made up. But it is a shorthand, not a description of what a receiver would actually output.
A GPS receiver does not trust its own clock. It solves for its position and its clock offset simultaneously, using an extra satellite to do it. A timing error shared identically by every satellite in view behaves like a receiver clock error, and a large part of it gets absorbed by that solve. If the only relativistic effect were a uniform, constant rate offset, the practical damage would be considerably less dramatic than the headline figure implies.
The reason relativity is nevertheless unavoidable is subtler and more interesting. GPS orbits are not perfectly circular. As a satellite moves through its slightly elliptical orbit its speed and its height both vary, so its relativistic clock rate varies too — by tens of nanoseconds, worth on the order of ten to fifteen metres of range error, and differently for each satellite at each moment. That is not a common-mode offset. It cannot be absorbed into a single clock-bias term, and receivers apply an explicit eccentricity correction for it. There is a further correction, the Sagnac effect, for the fact that the Earth rotates underneath the signal while it is in flight.
So the popular claim reaches a correct conclusion — relativity is genuinely load-bearing here — by an argument that does not quite hold up. The honest version is better: the fatal problem is not the big uniform drift, which the maths partly hides, but the small per-satellite variations, which it cannot. The standard technical reference is Neil Ashby’s Relativity in the Global Positioning System, published in Living Reviews in Relativity, which works through the corrections properly.
Why It Still Matters
Two reasons, and the second is the one people miss.
The first is that this is the largest continuously operating test of general relativity ever conducted. Not an experiment in a laboratory over a weekend — a system in constant use for decades, by billions of people, whose failure mode would be immediate and obvious. It has not disagreed with a theory published in 1915.
The second is that GPS is not primarily a navigation service anymore. It is a global time distribution service that also does navigation. Mobile network base stations use it to stay synchronized. Financial exchanges use it to timestamp trades. Electricity grids use it to keep phase measurements aligned across hundreds of kilometers. Data centers use it to order events.
That makes the constellation a single point of dependency for a great deal of infrastructure that has nothing to do with knowing where you are — which is why jamming and spoofing of satellite timing signals, now a routine feature of several conflict zones and increasingly of civil aviation near them, is a much larger problem than a few aircraft with confused maps.
The relativistic correction is the elegant part. The dependency is the part to worry about.