How accurate is AR measurement, really?

The short answer

In good conditions, over short distances, a phone AR measurement is typically within a centimetre or two of the tape. That holds for a wall of two or three metres in a bright room, with slow, steady handling and a properly calibrated session. Move away from those conditions — dim light, a plain wall, a long span, a hurried start, a fast pan that loses tracking — and the error grows, sometimes to several centimetres, and the app will not necessarily tell you that it has.

So the honest answer is: accurate enough for planning, layout and estimating; not accurate enough for cutting, fabrication or anything where a centimetre costs money. The rest of this guide explains what decides which side of that line a measurement lands on, and how to find out for your own phone.

What decides the error

Every phone AR measurement is the distance between two points the app has placed in a three-dimensional model of the room. Two things can go wrong: where the app thinks the points are, and how well the model matches the real room. Five factors dominate.

Tracking drift

The app works out where the phone is by watching visual features move through the camera image and fusing that with the accelerometer and gyroscope. Every estimate carries a small error, and those errors accumulate as you move. That accumulation is drift. Over a few seconds and a couple of metres it is small; over a long session with a lot of walking it is not, and a second point placed a minute after the first can be displaced by everything that happened in between. How AR measurement works covers the mechanism.

Lighting

Dim light means a noisier, blurrier image with fewer usable features, so tracking is less certain and the app’s estimate of scale can wander. Low direct sun is its own problem: it overexposes part of the frame and casts hard shadows that move with you.

Surface texture

Camera-only tracking needs visual detail. A patterned rug, brickwork or a bookshelf are ideal; a freshly painted white wall, a glossy tiled floor, a mirror or a glass partition are close to invisible. Point the phone at a blank wall for long enough and tracking degrades, and the app may place your point on the wrong surface or at the wrong depth.

Distance

Error grows with the distance from the phone to the point being placed, because a small angular error at the phone becomes a larger positional error further away. It also grows with the distance between the two points, because more movement means more drift. A 50 cm object at arm’s length is measured far more reliably than a five-metre wall from the middle of the room.

Depth sensor or camera only

This is the factor that separates devices rather than rooms, and it gets its own section.

Depth sensor versus camera only

A camera-only phone has to infer depth from motion: it sees the same feature from two positions and triangulates. That works well on textured surfaces close by and poorly on plain surfaces further away, and it cannot judge depth at all until the phone has moved.

A phone with a depth sensor — LiDAR on Apple’s Pro devices, a time-of-flight sensor on some Android phones — measures depth directly by timing light. It gets a depth map immediately, on plain walls and in dimmer light, and it can tell a wall from the floor before you have moved. That makes plane detection faster and more reliable, and the point you tap lands on the surface you meant rather than being estimated from the feature points nearby.

The practical difference is consistency rather than headline precision. On a good day in a textured room, a camera-only phone and a LiDAR phone can produce similar numbers. On a plain wall in a north-facing room, the LiDAR phone still produces a usable measurement and the camera-only phone may not. Which iPhones and iPads have LiDAR lists the devices.

What LiDAR and time-of-flight change, and what they do not

Apple’s LiDAR scanner is short range — Apple describes it as working up to about five metres — and its depth map is coarse compared with the camera image. That is enough to pin down surfaces and scale, but the fine position of a tapped point still comes largely from the camera. It does not turn a phone into a surveying instrument; a laser distance meter that measures a single point directly is still more precise over a long span.

Time-of-flight sensors on Android phones work on the same principle and bring similar benefits, with the details varying by model. ARCore also provides a depth estimate from motion alone on phones without a sensor, which helps with plane detection and occlusion but is less accurate than measured depth.

Realistic expectations by situation

Vertical measurements such as ceiling height are usually a little worse than horizontal ones of the same length, because the phone has to tilt through a range where the floor and ceiling offer few features to track.

How to test your own phone

You do not have to take anyone’s word for it, including this guide’s. Ten minutes with a tape gives you a number for your phone, in your rooms.

1. Lay a metal tape on the floor, extended to exactly two metres, in a well-lit room with a patterned or textured floor.
2. Open the app, calibrate slowly for ten seconds, then measure from the tape’s hook to the 2 m mark. Write the result down.
3. Repeat five times, restarting the session each time. Note the spread as well as the average.
4. Do the same at four metres, then along a plain wall, then with the lights off.
5. Finally measure a full wall with the phone and then with the tape.

The spread tells you more than the average. A phone that reads 1.99, 2.01, 2.00, 2.02 and 1.99 is one you can trust in that room; one that reads 1.96, 2.04, 2.00, 1.93 and 2.05 is telling you to measure twice and use a tape for anything that matters. Repeat the test occasionally — a new case, a scratched lens or a drop can change it.

When not to trust it

Use the phone for floor plans, furniture, rough material quantities, listings and first quotes. Reach for a tape or a laser measure when the measurement is for cutting, fabrication or made-to-measure ordering; for anything structural or load-bearing; for anything involving gas, electrics or fire safety; and for any purchase where a centimetre out means a return. Phone measurements are estimates. The tape is the truth you check them against.

Frequently asked questions

Is the iPhone Measure app accurate?

To the same degree as any phone AR measurement: typically within a centimetre or two over a short distance in good light, and less reliable over long spans, on plain surfaces or in dim rooms. On a LiDAR iPhone it is more consistent. It is a planning tool, not a replacement for a tape.

Is a LiDAR iPhone more accurate for measuring?

It is more consistent. LiDAR measures depth directly, so surfaces are found faster and more reliably, especially on plain walls and in dimmer light. It does not eliminate tracking drift, and it is still short range, so a long wall still deserves a check against a tape.

Why did the measurement change after I moved the phone?

The app is continually refining its model of the room, and when it gets better information it may re-place points slightly. A shift of a few millimetres is normal. A large jump usually means tracking was lost and re-established somewhere slightly different; remeasure after that happens.

Can I order a kitchen or blinds from a phone measurement?

Use the phone to plan and to get a quote, then confirm every dimension you are ordering to with a tape or a laser measure. Made-to-measure items, worktops and anything structural or safety-critical need a physical measurement.

Estimates, not certificates: every phone-based AR measurement is an estimate, TritonMeasure included. Verify anything safety-critical, structural or expensive with a physical tape or laser before you cut, order or sign.