Augmented Reality (AR) is rapidly maturing, and as this it develops, it will present the training community with unprecedented capabilities. However, there remain several key technical challenges that must be addressed in order to deliver credible and immersive experiences. At the Training & Simulation Industry Symposium 2019 and in subsequent public forums, the Simulation & Training Technology Center identified Dynamic Occlusion as a key technological challenge critical to the successful implementation of AR applications and specifically critical to AR training. Dynamic Occlusion is the ability of an AR system to realistically spatially integrate synthetic and real-world content. Dynamic occlusion allows moving real-world objects, such as people, to occlude virtual content in a credible and natural manner. For example, if a person walks in front of a synthetic box, the person should occlude the box and the box should not be visible “through” the person. This concept is intuitive to observers, but there currently does not exist a commonly understood metric for measuring dynamic occlusion performance of various AR systems. In this paper, we propose a metric for assessing Dynamic Occlusion and break Dynamic Occlusion down into constituent factors of false positive and false negative occlusion. We will discuss the relative merits and challenges of these differing types of occlusion and their ultimate effect upon user acceptance and suspension of disbelief in AR experiences. We will also discuss recommended methodologies to gather Dynamic Occlusion metrics over a variety of conditions. The intent of this work is to provide a common conceptual framework that can facilitate the establishment of requirements as well as objective comparisons of performance across various systems. This framework will allow solutions to be accurately compared as we collectively tackle this key technical challenge to the implementation of AR training systems.