ChatGPT in Thematic Analysis: Can AI become a research assistant in qualitative research?

The release of ChatGPT in November 2022 gives rise to an emerging scholarship on applying generative AI (GenAI) in qualitative data analysis, yet this new area remains underdeveloped. This article aims to assess how GPT-4 can help with thematic analysis using a sample interview dataset provided by Lumivero. In doing so, this article proposes a Guided AI Thematic Analysis (GAITA), an adapted version of King et al.’s (2018) Template Analysis. This framework positions researchers as a reflexive instrument and intellectual leader while thoroughly guiding GPT-4 in four stages: data familiarization; preliminary coding; template formation and finalization; and theme development. Additionally, I propose the ACTOR framework, a simple approach to combining different effective prompting techniques when working with GenAI for qualitative research purposes. Findings reveal GPT-4’s capacity for grasping the data, generating codes, subcodes, clusters, and themes, along with its adaptive learning and interactive assistance in organizing unstructured data and developing trustworthiness. However, this model has some key limitations in terms of its restricted context window for processing large datasets, its inconsistent outputs requiring multiple prompt attempts, the need to move across workspaces, and the lack of relevant training data for qualitative research purposes.