This paper introduces a practical and replicable socio-technical method for refining qualitative interview guides through structured human–AI collaboration. ChatGPT was used to role-play participants, allowing researchers to test and improve theory-informed questions through a two-layer workflow: (1) designing structured prompts that elicit realistic role-based responses, and (2) systematically evaluating those responses using criteria adapted from the Interview Protocol Refinement (IPR) framework. Applied in a healthcare study, this method improved question clarity, relevance, and alignment with theoretical constructs, and the resulting guide later produced rich and reflective responses in real interviews. The main contribution is a replicable early-stage prototyping method that shows how structured prompt engineering can be used to refine qualitative instruments with transparency and efficiency when traditional piloting is difficult. Although demonstrated in healthcare, the approach is transferable to other Information Systems contexts where expert access is limited.