Human-AI Collaboration in Inductive Thematic Analysis

As generative AI becomes increasingly integrated into qualitative analysis, a central methodological concern is how researchers can use AI to support coding without weakening human interpretive authority, contextual judgment, and the protection of participants' voices. This study examines how experienced qualitative researchers used an Inductive Thematic Analysis GPT (ITA-GPT), a purpose-built AI scaffold informed by Braun and Clarke's (2006) and Naeem et al.'s (2023) thematic analysis procedures, to first generate inductive thematic analysis outputs and then critically adjudicate AI-generated codes. Three qualitative research experts independently used ITA-GPT to analyze interview transcripts, and their interaction logs, AI-generated outputs, tracked revisions, and reflections were analyzed using the Human-AI Collaborative Inductive Thematic Analysis framework. The findings show that ITA-GPT was most useful in organizing the early stages of analysis. It supported transcript familiarization, generated initial verbatim and gerund-based codes, grouped early patterns, and helped researchers move systematically from data excerpts to provisional themes. Its strongest contribution was at the data-proximal stage, where in vivo codes helped preserve participants' exact language and created a traceable foundation for analysis. However, ITA-GPT was less effective when interpretation required contextual depth, emotional nuance, and understanding of institutional realities. In such moments, ITA-GPT produced codes that were sometimes too literal or disconnected from the Ghanaian teacher education context. Human researchers therefore modified, deleted, rejected, or inserted codes to restore contextual meaning and preserve participants' voices. The study concludes that AI can strengthen qualitative analysis when designed around qualitative inquiry principles, but human researchers must retain interpretive authority over meaning-making. The codes are available at: GitHub: https://github.com/matthewnyaaba/Inductive-Thematic-Analysis. The tool is available here: https://chatgpt.com/g/g-Fs5wk2FsM-inductive-thematic-analysis.