There is a growing need for automated systems that can interview patients and consumers about their health and provide health education and behavior change interventions using natural language dialog. A number of these health dialog systems have been developed over the last two decades, many of which have been formally evaluated in clinical trials and shown to be effective. This article provides an overview of the theories, technologies and methodologies that are used in the construction and evaluation of these systems, along with a description of many of the systems developed and tested to date. The strengths and weaknesses of these approaches are also discussed, and the needs for future work in the field are delineated.