3D alignment of a virtual model and corresponding real-world object is provided. A user places virtual tags on the real-world object as viewed through an augmented reality device. By placing these tags at the positions where he perceives certain real-world landmark locations, this allows for correction of misalignments between virtual content on the AR display and the real world. In particular, the virtual model include virtual landmarks that correspond to the real-world landmarks. Thus aligning the virtual landmarks of the virtual model to the virtual tags provides accurate alignment of the virtual model to the corresponding real-world object.