Documenting fidelity to HIV prevention interventions is critical to ensure consistency in intervention implementation and necessary for measuring intervention exposure and, ultimately, outcomes. Significant variation from prescribed protocols or inconsistent implementation can jeopardize the integrity of evaluation research and render outcomes uninterpretable. There is increasing support for HIV prevention models targeting seropositive individuals designed to be delivered by physicians during clinic visits. Assessing fidelity to physician-delivered interventions that occur during clinical exams present unique challenges. This paper presents findings from various data sources designed to track intervention fidelity and exposure to the Partnership for Health intervention, a physician-delivered HIV prevention intervention implemented in an urban community HIV clinic. We present findings from chart abstraction data, patient surveys and exit interviews, and provider qualitative interviews. Lessons learned and recommendations for maximizing the accuracy and validity of fidelity assessment in future evaluations of HIV prevention interventions in primary care settings are considered.