Reflexive human–AI collaboration: Tracing the evolving epistemics of qualitative inquiry (2021–2025)

AI- and LLM-assisted qualitative analysis has expanded rapidly across disciplines, yet the field remains unevenly synthesized in terms of how tools are integrated into workflows and how methodological rigor, transparency, and governance are addressed. This study maps the evolution of AI in qualitative research from 2021 to 2025 using SciMAT science-mapping analyses of 1862 Scopus-indexed publications and researcher-defined temporal windows (2021–2023, 2024, 2025). The approach characterizes within-period thematic structure via strategic diagrams and assesses between-period continuity via overlap analysis. Results indicate three time-sliced patterns: (1) early experimentation and methodological hybridization (2021–2023), where AI entered qualitative inquiry through assistive functions such as transcription, coding support, and sentiment analysis, anchored by themes including Semi-Structured Interview and Qualitative Research ; (2) systematization and applied integration (2024), where clusters such as Interview , Patient Care , and ChatGPT reflect expanding NLP/LLM use for transcript analysis, coding, and focus-group simulation alongside heightened debates on reliability, validity, and interpretive control; and (3) institutional maturity and ethical rebalancing (2025), where themes such as Health Personnel Attitude , Students , Human , and Qualitative Analysis signal stronger emphasis on ethical governance, AI literacy in graduate training, and equity- and context-sensitive use. Overall, the findings suggest a shift from exploratory hybridization toward more protocolized, reflexive human–AI collaboration, while underscoring persistent needs for transparency, auditability, and shared validity standards to keep AI-enhanced qualitative inquiry human-centered and interpretively robust.