The first peer-reviewed process covering all six steps of thematic analysis with generative AI.
Earlier studies used AI for only fragments of thematic analysis, gave the AI no research context, and never informed it of the researcher’s methodological considerations. This paper addresses all three limitations. It contributes both researcher training in thematic analysis and the development of inputs to train AI — a complete, transparent workflow from familiarisation to conceptual model, validated by comparing GenAI outcomes against a published manual analysis of the same data set.
Research focus, context, theory, philosophy, questions, methodology, participants
Realness, richness, repetition, rationale, repartee, regal
Robust, reflective, resplendent, radical, relevant, righteous
Reciprocal, recognizable, responsive, resourceful
Define new concepts with theoretical conceptual clarity
Synthesise concepts into a theory-grounded framework
Each analytical object has published selection criteria, so AI outputs can be evaluated rather than accepted:
The AI is only as rigorous as the context you give it. Before any analysis, the researcher familiarises ChatGPT with: the research focus, problem and context; the theoretical and philosophical underpinnings; the research questions, objectives and aim; a methodology summary; participants’ profiles; the process of data collection; and the six steps of systematic thematic analysis themselves — then asks the AI to confirm its understanding before proceeding. This is what separates a defensible AI-assisted analysis from pasting transcripts into a chatbot.
No. The researcher supplies the theoretical framing, guides every step, and retains interpretive responsibility. The AI’s contribution is speed, consistency, reduced human bias, and a transparent, auditable record of how the analysis was conducted.
Transparency is the key. Because this process documents every prompt and every step, it improves accountability compared with undocumented manual judgement. Researchers should follow their target journal’s AI disclosure policies and report the process openly.
The paper compares each step of a ChatGPT-conducted systematic thematic analysis against the corresponding manual analysis from a published study of scan-and-go shopping apps, using the same data set, research context, methodology and theoretical underpinnings.
It helps. The AI process follows the six steps of Systematic Thematic Analysis (Naeem et al., 2023), so reading that framework first gives you the methodological foundation the prompts are built on.