Handheld Augmented Reality (AR) is often presented using the magic-lens paradigm where the handheld device is portrayed as if it was transparent. Such a virtual transparency is usually implemented using video captured by a single camera rendered on the device’s screen. This removes binocular-disparity, which may undermine user’s ability to correctly estimate depth when seeing the world through the magic-lens. To confirm such an assumption this paper presents a qualitative user study that compares a magic-lens implemented on a mobile phone and a transparent glass replica. Observational results and questionnaire analysis indicate that binocular-disparity may play a significant role in participants’ depth perception. These promising results led to the subsequent implementation of a stereoscopic magic-lens prototype on a commercially available mobile device.