Augmented Reality overlays digital content on the real world using phones and increasingly head-worn displays.

What Is Augmented Reality (AR)? How It Works, Types, Devices & Real-World Uses (2025 Guide)

Updated October 2025

Augmented reality places 3D models, labels, and audio cues into your camera view or through transparent displays in real time. If you have tried an Instagram face filter or placed an IKEA sofa in your living room, you have used AR.

A Short History of AR

AR is a stack of ideas layered over decades. Milestones include Heilig’s Sensorama (1950s), Sutherland’s “Sword of Damocles” (1968), Krueger’s Videoplace (1975), the televised first-down line (1998), mobile AR via Wikitude (2008), and the global phenomenon of Pokémon GO in 2016. Today, AR is standardized on phones through ARKit and ARCore, and it reaches living rooms through mixed-reality passthrough on consumer headsets.

How Augmented Reality Works

AR does not replace reality. It adds to it. Under the hood, AR systems loop through sensing, understanding, anchoring, and rendering.

Sensing: Cameras capture frames while accelerometer, gyroscope, and magnetometer provide motion and orientation.
Understanding: Computer vision detects features, planes, and people. SLAM builds a map so content stays aligned as you move.
Anchoring: The app places objects on surfaces or specific images and keeps alignment consistent over time.
Rendering: A graphics engine like Unity with AR Foundation draws models with lighting, shadows, and occlusion.

What we observed in testing

On iPhone models with LiDAR, virtual object placement stabilized faster in low-texture rooms. In our room-walk tests across matte walls and plain floors, we saw noticeably less jitter and faster plane lock compared to non-LiDAR models.
Bright, even lighting reduced tracking drift. Harsh point lights created specular hotspots that confused plane detection and slowed initial placement.
On WebXR demos, 60 FPS was reachable on recent iOS and Android hardware when developers limited draw calls and used compressed textures. Excessive post-processing effects caused frame dips and anchoring wobble.

Core Computer-Vision Features

Plane detection: Finds tables, floors, and walls for object placement.
Image or marker tracking: Recognizes posters, QR codes, or packages to trigger overlays.
People occlusion: Hides parts of virtual objects behind people for realism.
Depth and LiDAR: Improves placement stability and physics.

Developers typically target ARKit (iOS), ARCore (Android), or WebXR for the browser.

Devices and Hardware in 2025

1) Mobile devices

iPhones and iPads, many with LiDAR, plus Android phones power the majority of AR experiences including try-ons, navigation, measuring tools, and education. To get started on iOS, see our How to use AR on iPhone.

2) AR headsets and smart glasses

Enterprise headsets: Hands-free workflows for remote assistance, training, and field service.
Consumer MR with passthrough: Select VR headsets provide high-quality color passthrough for spatial overlays used in design review and creative tools.
Lightweight smart glasses: Notifications, basic overlays, and voice or AI assistance.

Hardware components that make AR work

IMU sensors: Accelerometer, gyroscope, and magnetometer.
Cameras and depth: RGB, stereo, or LiDAR.
Compute: CPU, GPU, and neural accelerators for on-device inference.
Optics: Waveguides or reflectors in glasses.

Maker-minded readers can pair AR prototypes with physical parts. Start with our beginner 3D printer tutorial and avoid common errors with 3D printing mistakes.

The Four Core Types of AR

1) Marker-less (location or SLAM-based)

Uses GPS, IMU, and vision to understand surroundings without printed markers. Ideal for furniture previews or city overlays.

2) Marker-based

Triggers content when a specific image or object is recognized. Face filters are a special case where your face acts as the marker.

3) Projection-based

Projects light onto real surfaces such as exhibits or a “holographic” keyboard. Often paired with depth cameras for touch-like interaction.

4) Superimposition-based

Replaces or augments part of the scene with virtual content such as virtual try-ons or material previews.

Ready to browse gear ideas? See our Best AR gadgets.

AR Use Cases and 2025 Examples

Games and social

From Pokémon GO to city-scale hunts, AR blends fitness, exploration, and social play. Face-tracking effects show off the latest occlusion and segmentation tricks.

Education

Sky maps: Label planets and constellations in real time.
AR coloring: 2D drawings pop up as animated characters.
Translation: Overlay your language on signs and menus.

Navigation

Turn-by-turn overlays place arrows on streets in your camera view. In vehicles and aviation, heads-up displays project speed and guidance into the forward view.

Retail and try-on

Glasses, shoes, clothing, paint, or furniture previews reduce returns and speed decisions.

Healthcare and training

Surgeons practice with patient-specific 3D models. Technicians follow step-by-step overlays for faster and safer procedures. For clinical studies and guidance research, see IEEE Xplore.

Industrial and field service

Technicians view wiring diagrams and torque sequences on equipment. Remote experts annotate a live camera view. Warehouses use arrows to accelerate picking.

Exploring room-scale mixed reality at home? Our VR room buying guide covers space, floors, and cable management for safer sessions.

AR vs VR vs MR

VR immerses you in a fully digital world. The real world is out of view. AR keeps you in reality and layers content on top. Select devices support mixed reality by using color passthrough cameras and spatial mapping to blend virtual content with the room. For immersion fundamentals, see What is VR?

Want to Build AR? Start Here

Apple ARKit (iOS)
Google ARCore (Android)
Unity AR Foundation (cross-platform)
WebXR (web browsers)

Appendix: Our AR Testing Protocol

We run a consistent set of hands-on checks across iOS, Android, and select MR headsets to evaluate tracking stability, realism, and performance. Results inform the tips and observations you see in this guide.

Test Environment

Rooms: 1 low texture room with matte walls, 1 medium texture living room, 1 bright kitchen with reflective surfaces.
Lighting: 100–300 lux ambient, 500–700 lux task light, and 1 harsh point light to probe robustness.
Surfaces: Wood table, white desk, tile floor, and a patterned rug for plane detection variance.

Devices Under Test

Recent iPhone with LiDAR and a non LiDAR iPhone of the same generation.
Recent Android flagship and a mid tier Android phone.
One MR headset with color passthrough for room scale checks where relevant.

Core Procedures

Plane Lock Time: Measure seconds to first stable horizontal and vertical plane. Repeat 3 times per room. Report median.
Anchor Drift Walk: Place a 1 m virtual cube. Walk a 5 m rectangle around it and return. Record positional drift in centimeters.
Lighting Robustness: Repeat plane and anchor tests under ambient, task, and point light. Note failure modes like flicker or relock events.
Occlusion Realism: Use a human subject and a hand pass test. Score edge accuracy and temporal stability from 1 to 5.
WebXR Performance: For browser demos, log average FPS and 1 percent low using the onscreen stats display. Cap draw calls and use compressed textures for control builds.
Persistence: Close and reopen the app after 10 minutes. Check if anchors restore within 3 cm without manual realignment.

Scoring Rubric

Metric	Excellent	Good	Needs Work
Plane Lock Time	Under 1.5 s	1.5 to 3 s	Over 3 s
Anchor Drift (5 m walk)	Under 3 cm	3 to 8 cm	Over 8 cm
Occlusion Score	4 to 5	3	1 to 2
WebXR Avg FPS	58 to 60	45 to 57	Under 45
Anchor Persistence	Restores within 3 cm	3 to 8 cm	Over 8 cm or fails

Reporting Notes

All medians are calculated from at least 3 runs per scenario.
We record device model, OS version, app build, and lighting in lux for reproducibility.
Where applicable we include a short clip or screenshot sequence to illustrate failures or relock behavior.