This trips up nearly every AR project that starts with "it should just open from a link".
What works today:
Android, Chrome — WebXR with immersive-ar is properly supported. Plane detection, hit testing and anchors work. This is the good path.
iOS, Safari — WebXR immersive-ar is not available. If your plan is "iPhone users tap a link and get AR in the browser", that plan does not survive contact with iOS. Apple Vision Pro is the exception, not iPhone.
So for iPhone you have three honest options:
1. AR Quick Look with a USDZ file. Genuinely one tap, no install, works from a link or a QR code. The catch is that it is a viewer, not an app: you place a model and look at it. No custom logic, no UI of your own, no game.
2. A camera-based web AR library that does its own tracking rather than relying on ARKit through the browser. Works cross-platform, and image-target tracking is solid. World tracking is weaker than native and it costs money at any real volume.
3. A native app with ARKit. Everything works, and you have an install step between you and the user.
The decision that matters is not technical, it is what you are willing to trade. If the campaign lives or dies on "no install", accept the reduced capability. If the experience needs real interaction and persistence, accept the install.
The mistake we see most: a client is shown a WebXR demo on an Android phone, signs off, then discovers three quarters of their audience is on iOS.
What are you building? Happy to say which of the three it should be.
What works today:
Android, Chrome — WebXR with immersive-ar is properly supported. Plane detection, hit testing and anchors work. This is the good path.
iOS, Safari — WebXR immersive-ar is not available. If your plan is "iPhone users tap a link and get AR in the browser", that plan does not survive contact with iOS. Apple Vision Pro is the exception, not iPhone.
So for iPhone you have three honest options:
1. AR Quick Look with a USDZ file. Genuinely one tap, no install, works from a link or a QR code. The catch is that it is a viewer, not an app: you place a model and look at it. No custom logic, no UI of your own, no game.
2. A camera-based web AR library that does its own tracking rather than relying on ARKit through the browser. Works cross-platform, and image-target tracking is solid. World tracking is weaker than native and it costs money at any real volume.
3. A native app with ARKit. Everything works, and you have an install step between you and the user.
The decision that matters is not technical, it is what you are willing to trade. If the campaign lives or dies on "no install", accept the reduced capability. If the experience needs real interaction and persistence, accept the install.
The mistake we see most: a client is shown a WebXR demo on an Android phone, signs off, then discovers three quarters of their audience is on iOS.
What are you building? Happy to say which of the three it should be.