<p>Generative artificial intelligence is reshaping analytical practices and raises questions about how meaning is constructed when analysts work in dialogue with a generative system. This article presents <i>A Duo with ChatGPT</i>, a structured instrument designed to make hybrid interpretation empirically traceable during a qualitative analysis task. Our contribution is methodological, focussing on instrument design and feasibility rather than evaluating ChatGPT’s performance. The framework provides an operational definition of hybrid interpretation and specifies four epistemic dimensions (distributed agency, interpretive abduction, epistemic control, and analytic metacognition). The instrument organises the task into successive phases that alternate between human-only interpretation and guided interactions with ChatGPT and asks participants to record prompts, system outputs, decisions, and brief annotations of significant moments. The instrument was piloted with three participants with education-related backgrounds and different profiles, an undergraduate student in Pedagogy, an in-service primary teacher, and a doctoral researcher. The pilot suggests that the instrument is feasible and generates analysable process records that can be read through the four epistemic dimensions in participants’ written accounts. The article discusses how these traces support a process-based reading of hybrid interpretation and outlines directions for refinement and replication.</p>