Comparing Use of Human-Created to AI-Generated Instructional Materials to Teach Science Vocabulary: A Randomized Controlled Trial

Evidence-based vocabulary instruction is essential for literacy and content learning, yet many educators face barriers, such as lack of training and time constraints, to implementing it. Content Acquisition Podcasts (CAPs) are multimedia instruction that incorporate evidence-based practice to teach vocabulary yet can be arduous to create. This randomized controlled trial (RCT) examined whether generative AI-created Content Acquisition Podcasts–Student (CAP-S) produced vocabulary outcomes comparable to researcher-created CAP-S and whether cognitive load predicted performance. Participants included 116 fifth- and sixth-grade students, including students with disabilities (SWD), randomly assigned to AI-generated or researcher-created CAP-S across three science vocabulary lessons. No statistically significant differences in vocabulary outcomes were found between conditions at any assessment point in the study. Results also showed that perceived cognitive load significantly predicted performance, while instructional condition did not. Findings suggest AI-generated CAP-S may provide a scalable approach, that also mitigates workload burdens of educators, without compromising instructional effectiveness.