iPhone LiDAR Accuracy in 2026: What the Peer-Reviewed Data Actually Says

How accurate is the iPhone's LiDAR scanner? For objects larger than 10 cm: about ±1 cm. For surfaces within 0.3-3 m: about 1-2 cm. At building scale: 10-20 cm. Those are peer-reviewed numbers, not marketing - and each comes with conditions that matter more than the headline figure.

This post collects the measured data in one place, because most of what ranks for this query is either a spec sheet or an opinion. Accuracy on this sensor is conditional on distance, surface, lighting, and - critically - the software pipeline of the app doing the scanning. All four are quantified below.

What the sensor actually is

The LiDAR scanner in iPhone Pro and iPad Pro models is a direct time-of-flight (dToF) sensor: a VCSEL infrared laser array that measures true hardware depth, fused with the camera stream and the inertial sensors through visual-inertial odometry. That fusion is why it initializes instantly, works in darkness, and handles blank white walls that defeat camera-only AR - and also why its errors are systematic and worth understanding rather than random noise.

How accurate is iPhone LiDAR, by scenario?

Two readings of this table. First, at close range on cooperative surfaces, the sensor is genuinely centimeter-class - good enough that clinical teams publish validation studies instead of dismissing it. Second, error grows with scale: a room scan and a building scan are different measurement problems, and anyone quoting one number for both is selling something.

What are the hard limits no app can engineer around?

Four constraints come from physics and Apple's power budget, not from software quality:

• Range: about 5 meters. That is Apple's official figure, capped by what a phone battery can drive through a laser array. In practice this blocks reliable exterior capture of anything above the first story from ground level.
• Material failures. Glass, mirrors, polished metal, dark or transparent surfaces, and liquid boundaries scatter the laser or let it pass through, returning false depth readings placed behind the real surface. Direct sunlight adds infrared interference.
• Fine detail. Point spacing runs from several millimeters to about a centimeter, so thin objects - cables, pipe bends, bolt heads, table legs - blur or vanish from the mesh.
• Scene size. Apple's RoomPlan API is room-centric by design: recommended rooms up to about 9x9 m, at least 50 lux of light, and sessions kept under about five minutes before thermal and battery limits bite.

None of these are dealbreakers for a well-scoped product. All of them are dealbreakers for a product scoped as "a laser scanner in your pocket."

Why do two apps on the same iPhone give different results?

The most underappreciated number in the literature: an independent benchmark scanned the same scenes with the same hardware through two popular scanning apps and measured a 5 cm mean deviation (10 cm RMSE) for one and 44 cm (56 cm RMSE) for the other. Identical sensor, roughly 9x difference in output quality.

The explanation is that raw LiDAR frames drift. The phone's pose estimate accumulates error as you move, and everything downstream - registration, loop closure, outlier filtering, mesh reconstruction - either corrects that drift or bakes it in. The software pipeline dominates the outcome, which means "iPhone LiDAR accuracy" is really a property of the app, bounded below by the sensor.

We have lived this from the builder's side. Our team ships a production scanning app for a surveying technology company: iPhone LiDAR capture fused with an external RTK GNSS receiver over Bluetooth, streaming correction data from an NTRIP base-station network, with a per-frame constraint engine that rejects frames on velocity, movement delta, and GNSS fix quality before they ever enter the point cloud. The result is sub-5 cm georeferenced accuracy in field conditions, with a live preview rendering up to 15 million points at 60fps. The sensor was never the bottleneck. The pipeline - drift handling, quality gating, georeferencing - was the entire project.

Can an iPhone replace a terrestrial laser scanner?

No, and products that claim otherwise create expensive disappointments. Survey-grade terrestrial laser scanners deliver millimeter-class accuracy at ranges the phone cannot reach, and dedicated fixed dimensioning stations carry certifications a phone cannot hold.

The honest framing is different: centimeter-grade, in minutes, in every field worker's pocket - against millimeter-grade, in hours, from one expensive device. For first-pass documentation, scoping, claims evidence, floor plans, and pre-survey inspection, the first profile wins on economics. For structural fabrication tolerances or legally certified measurement, it does not, and a serious product says so in its own marketing.

What this means if you are evaluating LiDAR for a product

Three practical conclusions from the data:

1. Match the accuracy tier to the use case, in writing. Centimeter-class at close range covers claims documentation, floor plans, dimensioning, and clinical measurement. Building-scale as-builts at 10-20 cm are a scoping tool, not a fabrication reference.
2. Budget for the pipeline, not the capture. Apple's APIs make capture nearly free; the 9x spread between apps shows where the engineering actually lives. Drift correction, quality gating, and export semantics are the product.
3. Benchmark on real sites before committing. Twenty test scans against a tape measure and a laser rangefinder on your actual scenes will tell you more than any published study - including this one - because your surfaces, lighting, and scan discipline are the conditions that matter.

We map where this sensor is earning money across industries - construction, insurance, logistics, healthcare - in the 2026 industry map. And if you are weighing a LiDAR feature or product and want the accuracy question answered for your specific use case, that conversation is what we do - see our AR and spatial engineering services.