Separating the place where you build AR content from the place where it appears is the root of most alignment errors.
There is a quiet assumption built into most outdoor augmented reality workflows. It says that the place where you create content and the place where that content appears can be separated, as long as you write down the coordinates carefully enough.
That assumption is where the trouble starts. A developer stands on a footpath with a phone, notes a latitude and longitude, walks back to a desk, opens Unity with an AR SDK, and builds a virtual sign. Weeks later the sign hovers half a metre to the left of where it belongs, or floats above the shopfront instead of beside the door.
Desktop tools work from fixed or prerecorded environmental data. They cannot see that the tree grew, that a delivery van now parks there, or that the pavement was repainted last spring. Every one of those changes shifts the visual features the tracking system relies on, so the content drifts. The team behind a 2025 study in Electronics names this as the central limitation of the mainstream approach.
Their answer was to move authoring into the moment and the location. Stand at the point of interest, create the object, store it, and let the localization happen live. The reasoning is straightforward: if the anchor is captured under the same conditions the user will encounter, far less has to be corrected later.
None of this makes desktop tooling obsolete. Complex 3D assets still belong on a workstation. But the placement step, which is where accuracy is won or lost, now has a credible alternative that fits in a pocket.
Brata, K.C., Funabiki, N., Panduman, Y.Y.F., et al. (2025). A Proposal of In Situ Authoring Tool with Visual-Inertial Sensor Fusion for Outdoor Location-Based Augmented Reality. Electronics.
