VRHow / XR Technology / What is augmented reality?

What is augmented reality?

Short answer: Augmented reality (AR) adds computer-generated information—such as labels, a 3D model or an arrow—to your view of the physical world. The useful part is not merely drawing graphics over a camera image; it is keeping those graphics aligned with the place, surface or object they describe.

Last checked October 11, 2026. This is an evidence-led explainer, not a hands-on device review. Device support, tracking quality and features vary by app, hardware, software version and region; VRHow has not independently tested the products discussed.

AR is an overlay with a relationship to the real world

Seeing a virtual hat in a selfie, a repair arrow attached to a machine, or a sofa previewed at room scale are all AR-style experiences. The display may be a phone, a see-through headset or a camera-based headset. What unites them is that the real environment remains visible while digital content is added to it. Microsoft describes overlays of graphics, video streams or holograms in the physical world as augmented reality, while NIST defines AR as technology that overlays digital information on an image of something being viewed. Those definitions are useful, but not every “AR” label promises the same level of spatial alignment. [NIST]

That distinction matters. A floating sticker that follows the screen is an overlay, but it is not the same technical achievement as a virtual measuring guide that stays on a wall while you walk around it. In ordinary conversation, both may be called AR. For a buying or design decision, ask what the content is anchored to and how it behaves when lighting, surfaces or the camera view change.

How an AR system makes an object look “there”

Most mobile AR follows a short, constantly repeated pipeline:

Sense motion and features. Cameras look for distinctive visual features across frames; inertial sensors measure movement. ARCore calls this simultaneous localization and mapping (SLAM), combining camera features and IMU data to estimate the camera’s position and orientation. [Google ARCore] Apple describes the comparable ARKit process as visual-inertial odometry. [Apple ARKit]
Understand enough of the scene. The software can infer planes such as a floor, table or wall, test where a screen tap intersects the scene, and—on supported hardware—produce depth information. ARCore notes that low-texture surfaces such as a plain white wall may be difficult to detect. [ARCore fundamentals]
Choose a coordinate system and anchor. An app places a model at a calculated pose, then attaches it to an anchor or trackable. As the system refines its estimate, the anchor helps preserve the model’s relationship to the environment rather than to a single display frame. [Google on anchors]
Render the view. The app redraws the virtual object from the camera’s changing viewpoint. Lighting estimates can adjust the object’s brightness and color; depth can help a real object appear in front of, or behind, virtual content. These are estimates, not a guarantee that the digital object is physically understood like a human would understand it. [Google on depth and light]

Head-worn AR uses the same broad idea, but the display and input change. A see-through display lets light from the world reach your eyes; a camera-based system shows a live camera feed on an opaque display. That difference affects latency, contrast, field of view and how confidently you can judge the real scene. It is one reason “AR” is a category, not a single product experience.

What AR can do—and what it cannot promise

Capability	What it can mean in practice	Important limitation
World tracking	Content appears to stay put as you move.	Tracking quality can fall in darkness, blur, rapid movement or visually empty scenes. Apple documents these conditions.
Plane detection	A virtual object can be placed on a detected floor, table or wall.	Detection is an estimate that may improve or shift over time; unusual, shiny or textureless surfaces are harder.
Depth and occlusion	A real chair or person may hide part of a virtual object.	Depth is device- and API-dependent, and errors at edges or on moving objects are possible.
Image recognition	A poster, package or marker can trigger content.	The app must recognize the intended image; angle, distance, lighting and obstruction matter.

So AR is not X-ray vision, perfect measurement or a substitute for safety judgment. A navigation arrow can be visually helpful without being a certified instruction. A furniture preview can show approximate placement without proving that the item will fit through a doorway. Treat spatial registration as a confidence level to verify, not as a physical guarantee.

AR, VR and mixed reality: a practical boundary

Our VR-versus-AR-versus-MR guide gives the quick comparison: VR replaces your view; AR keeps the real world visible and adds information; mixed reality is often used for experiences in which digital objects understand and respond to the surrounding space. In practice, the labels overlap. A phone AR filter and a headset with room-aware occlusion are both called AR, even though their sensing and interaction are very different. Judge the behavior—camera view, display type, tracking, depth, anchors and input—rather than the marketing term alone.

How to decide whether an AR feature is genuinely useful

Before choosing an app, headset or workflow, ask:

What must remain aligned? A cosmetic overlay may tolerate drift; a measurement, assembly or medical visualization needs a much stricter error policy.
What environment will it face? Test the actual lighting, surfaces, movement, people and reflective materials. Do not infer performance from a controlled demo.
What happens when tracking degrades? A responsible experience should show a tracking warning, pause placement or provide a safe fallback—not silently present stale guidance.
What data leaves the device? Camera and spatial data can be sensitive. Read permissions and retention terms, especially for workplace or home mapping.
Can a person use it safely? Keep awareness of traffic, stairs, tools and other people. NIST’s AR usability framework emphasizes evaluating effectiveness, efficiency, satisfaction and safety in the specific context of use, rather than assuming a visually impressive prototype is usable. [NIST usability framework]

The best AR is therefore not the one with the most dramatic holograms. It is the one that adds information at the moment it helps, stays honest about uncertainty, and fails in a visible and safe way when its view of the world is incomplete. For web-based experiences, the W3C WebXR specification distinguishes an immersive-ar session from VR, but browser and device support still determines whether a particular experience can run.

Sources and method

Sources were opened and checked on October 11, 2026. Platform documentation describes capabilities, not a promise that every phone or headset supports them. The NIST source is a 2022 research framework, so its human-factors guidance is more general than any single product’s feature list.