An augmented-reality experience can lose its audience before the camera opens. A person follows a link, reaches an app-store page, waits for a download and then has to find the intended content again. For something as small as viewing one object, that is a large amount of setup.
Betty Peng's Web Powered Augmented Reality talk points toward a more direct route. Its examples and prototypes explore how the web can deliver AR experiences, including work around Google I/O. The attractive part is the distribution model: a useful experience begins from an ordinary web address.
The browser doesn't remove the underlying tracking problem. An image-marker demo, a face effect and an object anchored to a room all need different information. Calling them WebAR doesn't make their requirements interchangeable. A prototype should state which relationship it can maintain: attached to a detected image, attached to a face, or placed in a tracked physical space.
That choice shapes the interaction. A marker supplies a visible instruction because the visitor can point the camera at it. A room-tracking experience may require an initial scan or enough texture in the environment. If the page gives no feedback during that stage, the visitor cannot tell whether they should wait, move the device or try a different surface.
Browser capabilities also vary. In this period, experimental AR paths and device-specific integrations should be treated as such. A demonstration working in one configuration doesn't imply that every phone opening the URL can reproduce it. The page needs to detect what it can use and offer a sensible alternative when the required capability is absent.
For a product model, that alternative might be a conventional 3D viewer. The visitor still gets the object, can inspect its shape and can read the information that motivated the visit. AR then adds spatial context where available. A blank screen with an unsupported message gives the content no second chance.
Download size is part of the entry experience too. A browser removes an installation step, but a large model and texture set still have to arrive. Loading progress, a lightweight preview and a clear camera-permission explanation can make the wait comprehensible. None is technically glamorous, yet all influence whether the person reaches the spatial part.
I like the web for experiments whose value can be understood quickly: place one object, reveal one relationship, compare one scale. That scope fits the casual commitment implied by a link. A long application can justify a more involved setup, but a brief encounter should earn each additional step.
The interesting promise is therefore a shorter path between curiosity and experience. The tracking still has to work, and the interface still has to explain its limits. The browser gives that work a more accessible front door.