How accurate is iPhone GPS?

About five metres outdoors with an unobstructed view of the sky. Indoors, in a car park or between tall buildings it is routinely twenty to a hundred metres, and the useful part is that your phone usually knows which of those it is giving you.

People tend to think of GPS as either working or not working. It is better understood as a confidence level that changes second by second, and iOS tells you what that level is if you know where to look.

The number nobody reads

Every position iOS hands an app carries an accuracy radius in metres. It is a statement of uncertainty: the phone believes you are somewhere within that circle. It is not a margin of error around a known point, because there is no known point.

This is why the blue circle in Maps grows and shrinks. A large circle is not the phone being wrong, it is the phone being candid.

The field is called horizontalAccuracy. Apple describes it as a radius of uncertainty in metres, but does not publish the confidence behind it, so there is no documented figure for how often the true position falls inside that circle. Read it as the phone’s own estimate of how it is doing, not a guarantee.

A negative value in the same field means the latitude and longitude are not valid at all: the difference between a bad fix and no fix. Apps are supposed to check for it. Not all of them do.

Vertical accuracy is reported separately and it is worse. Satellite geometry is far better at placing you on a map than at placing you above it.

What you actually get

Those are typical ranges from ordinary use, not a controlled survey. We have not stood in a hundred places with a reference receiver and measured the error, and neither has anybody quoting one authoritative figure. The spread is the honest answer: the same phone will give you five metres on one street and eighty on the next.

It was never only GPS

The phone does not contain a GPS chip in the way people picture one. It has a location stack, and satellites are one input into it. Recent iPhones receive several constellations rather than only the American GPS one, with GLONASS, Galileo, QZSS and BeiDou listed in Apple’s specifications. More constellations means more satellites overhead, which matters most where half the sky is behind a building.

Alongside that, iOS positions you from the Wi-Fi networks in range, from cell towers, and from the accelerometer and gyroscope, which can carry a position forward briefly when the satellites drop out. What comes out of the far end is one blended answer with one accuracy number. Apple does not publish how the blend is weighted, so when a position is wrong there is no way from outside to say which input caused it.

The first fix is the worst one

Open Maps after a flight and the dot lands somewhere approximate before snapping to where you actually are. That is not the phone being slow, it is the phone doing something specific.

A receiver has to know where the satellites currently are before it can use them, and that data is broadcast slowly by the satellites themselves. Phones fetch it over the network instead, which is why a fix arrives in seconds on Wi-Fi and takes longer with no data connection at all. Until it lands, what you are looking at is Wi-Fi and cell positioning wearing the same blue dot.

So the first ten or twenty seconds after you open a location-aware app are the least accurate it will ever show you. Give it a moment.

Why cities are the hard case

GPS works by timing. The receiver measures how long a signal took to arrive and derives distance. In a city, a signal frequently reaches you after bouncing off a glass facade, so the measured path is longer than the true one and the position is pulled sideways.

This is called multipath, and it is the reason your phone occasionally places you confidently on the wrong side of a street you are standing on.

It also means the city penalty is not a fixed amount. The same corner can be accurate facing one way and thirty metres out facing the other, depending on which building the signal bounced off. A phone that is reliably wrong in one particular place is a different problem.

Precision is not accuracy

A coordinate printed to seven decimal places looks authoritative. Seven decimal places of latitude is roughly a centimetre, and no phone knows where you are to a centimetre. The digits are free: they fall out of the arithmetic rather than the measurement, and a position can be reported to eight places while sitting ninety metres from the truth.

Navigation apps add a second layer. They do not plot the raw coordinate, they snap it to the nearest plausible road, because a dot skating through the gardens beside a motorway is worse for a driver than a dot in the wrong lane. The dot on screen and the coordinate the app received are not the same thing.

Things that quietly make it worse

Most everyday degradation is not the satellites. It is configuration, and it is on your phone:

• Precise Location off for an app. iOS hands that app a broad area rather than a point, refreshed only occasionally. We have not measured how broad, and Apple’s own description is deliberately unspecific.
• Wi-Fi switched off. Indoors that removes the input doing most of the work.
• Low Power Mode. It does not switch GPS off, but background activity is trimmed, so a location may be older than you assume.
• A pocket, a thick case, a metal roof. Anything between the antenna and the sky costs signal, and a car is a metal box with windows in it.

It is never completely still

Leave a phone on a table and watch the dot. It wanders a few metres. Every fix is an independent estimate with its own noise, and nothing is averaging them into a fixed point.

Worth knowing if you are interested in how location gets scrutinised: a position that never drifts at all is the one that looks unusual. Real GPS is never perfectly still.

Why any of this matters to you

Almost nobody arrives at a page about GPS accuracy out of interest in satellites. They want to know why an app placed them somewhere they were not, or how much to trust a location an app showed them about somebody else. Both questions have the same unsatisfying answer: one coordinate is not evidence of much. It carries an uncertainty the app may never have shown you, and it may have been the first fix of the session rather than a settled one.

The one place this touches what we build: a held position does not accumulate error the way a live satellite fix does, which is the point of the section above. We have not measured what accuracy radius iOS attaches to a simulated position, on any iOS version, so we are not going to claim one here. What apps can and cannot infer from a location has its own guide.