The Use of ChatGPT in Qualitative Data Analysis: A Comparative Analysis on Contextual Fidelity and Thematic Consistency

Authors : Serdar Erdem, Akif Pamuk

Pages : 28-57

View : 71 | Download : 53

Publication Date : 2025-08-05

Article Type : Research Paper

Abstract :This study aims to evaluate the functionality of the large language model ChatGPT in qualitative data analysis by comparing it with human-conducted content analysis. The dataset consists of a total of 43,164 tweets shared on the X (formerly Twitter) platform on Teachers’ Day, November 24, in the years 2020 and 2021. Since the data were obtained from previously archived social media content, document analysis was used as the data collection method. The research was designed as a comparative case study. During the analysis process, both descriptive analysis and content analysis techniques were employed. The dataset, previously used in a published study by the researcher, was uploaded to ChatGPT-4o for AI-assisted analysis. The findings were compared in terms of thematic consistency, contextual fidelity, conceptual depth, and example discourse alignment. Additionally, ChatGPT’s critical evaluations and theoretical framework suggestions regarding the dataset were also assessed. The results indicate that while ChatGPT is effective in constructing conceptual frameworks and generating general themes, it exhibits limitations in understanding context and producing data-faithful examples. Particularly in the absence of user guidance, the model tends to generate synthetic discourse, posing a risk of contextual drift. Therefore, when supported with researcher oversight and clear prompts, ChatGPT can serve as a complementary tool in qualitative analysis processes.
Keywords : ChatGPT, Yapay Zekâ Destekli Nitel Analiz, Yapay Zekâ ve Bağlam Sadakati, ChatGPT ile Tematik Kodlama, Karşılaştırmalı Durum Çalışması

ORIGINAL ARTICLE URL