I don’t have an Android so I can only speculate, but it appears the augmented browsing is not real-time. So unlike Layar/Wikitude when you move your device around the various points of interest are plotted in the camera view, Goggles requires you to take a picture, the 2D image is then send to the server to be recognised so it gets around the problem of having to deal with complex 3D shapes, distance etc

The common thing between pictures in a gallery, books and landmarks, is that they are relatively 2d. You mostly see and take pictures of landmarks from relatively far away, where their 3d structure is relatively lost.
Cars in the street, on the other hand, are a whole different story. I wonder how good it is in recognizing famous (non land-markian) sculptures, for example Venus de Milo.