Thematic analysis involves six phases: 1) data familiarisation, 2) code generation, 3) theme identification, 4) theme review, 5) theme naming, and 6) theme interpretation. Manually, analysing data from large corpora of discourse is often time-consuming, delaying feedback and potentially impairing the researchers. Fortunately, the rise of AI-powered tools like Google's NotebookLM is helping ease this burden. Despite the growing adoption of LLMs in qualitative research, there remains a lack of practical and replicable frameworks to conduct research, particularly for beginners.

This study aimed to provide the desired framework using the ADDIE model through modeling and testing of Zhang et al’s (2024) prompt engineering framework for thematic analysis using ChatGPT. By adapting and implementing the developed prompt along with a code book, and analysis database in cycles,  the researchers were able to analyse 12 types of figurative language By Perrine from 13 live sport commentary match transcriptions. After some curation, the result indicates 251 utterances applying variations of types of figurative language.

As derivation and repetition often occur during the coding process, the adoption of NotebookLM still requires researchers’ active participation. Nevertheless, the developed prompt did ease the time consuming manual labour.