Effective learning interventions (online courses, SIMS, live instruction, and self-directed activities) must be strongly aligned with instructional goals. Programs such as the Personal Assistant for Learning (PAL) being developed by the US Advanced Distributed Learning initiative and the Generalized Intelligent Framework for Tutoring (GIFT) developed by the Army Research Lab (ARL) emphasize the Government’s investment in learning interventions that adapt to learner goals and preferences. To be practical, such systems must automatically detect and align digital content and other learning intervention with learning goals. The research reported here addresses one step in this process. It is part of the larger integration effort between GIFT and Tools for the Rapid Development of Expert Models (TRADEM), supporting the efforts and goals of the Army Research Lab (ARL). This paper presents techniques that automatically use a set of text-based features to detect pedagogically appropriate topics. These techniques are part of an attempt to automate portions of the front-end anal-ysis and design steps in the tradition “ADDIE� (analysis, design, development, implementation, and evaluation) [Branson et. al., 1975] approach to content creation. This paper sets the context for this work, describes the tech-niques and algorithms used, and provides data that shows that auto-detection performs well when reviewed by and compared to hand-generated mappings by instructional design experts.