In community-based interventions involving lay health workers, or "community health workers," peer-client interactions are not typically observed by investigators, creating challenges in assessing intervention fidelity. In the context of a community-based randomized controlled trial of the effectiveness of peer support on diabetes outcomes of people with diabetes in rural Alabama, a region characterized by poverty and low literacy, we developed a video assessment tool that assessed participant perceptions of peer-client interactions. The video assessment consisted of four short skits on areas of emphasis during peer training: directive versus nondirective counseling style and setting a specific versus a more general goal. The video tool was evaluated for association with questionnaire-derived measures of counseling style and goal setting among 102 participants. For counseling style, 44% of participants reported that their peer advisor was most similar to the nondirective skit. For goal setting, 42% reported that their peer advisor was most similar to the specific goal skit. There was no statistically significant relationship between skit selection and questionnaire-derived measures. The video assessment was feasible, but results suggest that video and questionnaire assessments in this population yield different results. Further validation to better understand the differences between questionnaire reports and video assessment is warranted. © 2014 Society for Public Health Education.