Augmented reality (AR) technologies are one method of supporting military visual search tasks, such as target identification, recognition, and acquisition. However, whether or not a given AR technology actually improves human performance depends on many factors, including the quality of the display and the quality of the AR information provided to the Soldier. In this paper, we describe current research efforts by the U.S. Army CCDC C5ISR Night Vision and Electronic Sensors Directorate to use simulation to study one aspect of AR information quality: spatial accuracy. Specifically, we examine the level of AR spatial accuracy required to improve human performance as a function of the density of potential targets. Participants were placed in virtual scenarios and asked to locate and target a single virtual human holding a weapon amongst many unarmed virtual humans. Participants used realistic sensor controls to scan the virtual field of regard and to locate the target. Baseline performance was characterized by having participants locate targets without any AR assistance. In other control trials, participants were guided to the target with perfectly accurate AR symbology, located both on a situational awareness ring and in the operator’s field of regard. In experimental trials, participants were guided by imperfect AR symbology distorted by fixed amounts of angular error. These experimental conditions of varying AR information were crossed with various densities of potential targets (i.e., virtual humans) in the field of regard, ranging from densely-populated to sparsely-populated search areas. Our results examine the effects of AR spatial accuracy on target acquisition time, comparing imperfect AR to perfect AR and unaided searching in scenes with different densities of potential targets. The ultimate goal of our research program is to support product development and virtual prototyping by simulating task-specific and sensor-specific AR accuracy requirements for sensors and head-up displays.