Reflexive Human–AI Collaboration: Tracing the Evolving Epistemics of Qualitative Inquiry (2021–2025)
A
JovitoAnito
Jr.
1✉Emailjovito.anito@gmail.com1College of Teacher EducationJose Rizal Memorial State UniversityDapitan CityPhilippines
Jovito Anito Jr.
College of Teacher Education, Jose Rizal Memorial State University, Dapitan City, Philippines
jovito.anito@gmail.com
Abstract
This study maps the evolution of Artificial Intelligence (AI) in qualitative research from 2021 to 2025 using SciMAT analyses of 1,862 Scopus-indexed publications. Three phases emerged: (1) early experimentation and methodological hybridization (2021–2023), (2) systematization and applied integration (2024), and (3) institutional maturity and ethical rebalancing (2025). In the first phase, AI entered qualitative inquiry through assistive functions, mainly transcription, coding support, and sentiment analysis, primarily in health and social science research. Themes such as Semi-Structured Interview and Qualitative Research anchored this stage, reflecting efforts to merge computational efficiency with interpretive depth. By 2024, AI methods became routine in qualitative workflows. Clusters including Interview, Patient Care, and ChatGPT show how NLP and large language models supported transcript analysis, coding, and focus-group simulation while prompting debates on reliability, validity, and human interpretive control. By 2025, the field exhibited institutional consolidation. Major themes, such as Health Personnel Attitude, Students, Human, and Qualitative Analysis, signaled the rise of ethical governance, AI literacy in graduate training, and increased attention to equity and contextual sensitivity. AI was increasingly viewed as a reflexively managed collaborator rather than a replacement for human analysis. The findings reveal a clear trajectory from early hybrid experimentation to reflexive human–AI partnership. The study demonstrates how qualitative research is being reorganized technically, ethically, and pedagogically, and highlights the principles required to ensure that AI-enhanced inquiry remains human-centered and interpretively robust.
Keywords:
AI-assisted qualitative research
hybrid human–AI analysis
thematic evolution
large language models (LLMs)
reflexive integration
qualitative data analysis (QDA)
A
1. Introduction
This study is situated within the evolving relationship between qualitative inquiry and artificial intelligence. It first revisits the interpretivist and critical foundations of qualitative data analysis and the diversification of its analytic traditions, then traces the digital turn through CAQDAS and the emergence of AI-assisted methods. Building on this, it reviews how AI and large language models have shifted from pilot tools to hybrid, ethically contested partners in qualitative workflows. Against this backdrop, the study’s objectives and research questions focus on mapping how these developments are thematically configured, how they evolve over time, and how they re-shape the methodological, ethical, and pedagogical horizons of qualitative research.
1.1. Literature Review
Qualitative research has long evolved as an interpretive, context-sensitive enterprise seeking to understand how people construct meaning from lived experience and social interaction. Unlike quantitative paradigms focused on measurement and generalization, qualitative inquiry privileges depth, complexity, and subjectivity (Gillan et al., 2014; Whitley, 2008). Grounded in interpretivism and constructivism, it assumes that reality is socially constructed and best understood through participants’ perspectives (Galuppo & Benozzo, 2024; Goldkuhl, 2012).
Over ensuing decades, analytic traditions diversified. Thematic analysis remains a flexible method for pattern identification (Thompson et al., 2022; Williamson et al., 2018); content analysis extended categorization to varied media (Peake & Koleth, 2024); grounded theory sustained iterative theorizing (Pope et al., 2019); and narrative analysis explored meaning through storytelling (Liamputtg, 2009). Reflexivity has become a defining hallmark wherein researchers must continually interrogate their assumptions and positionality to preserve transparency and ethical integrity (Dodgson, 2019; Namatovu & Langevang, 2025; Olmos-Vega et al., 2023; Sarfo & Attigah, 2025; Sirris, 2022).
1.1.1. Artificial Intelligence and the Emergence of AI-Assisted Qualitative Analysis
Digital innovations reconfigured qualitative practice. The rise of Computer-Assisted Qualitative Data Analysis Software (CAQDAS) revolutionized data management, enabling efficient handling of complex datasets while maintaining analytic rigor (Kikooma, 2010; Ntsobi et al., 2024). These developments laid the groundwork for the next methodological transformation—AI-assisted qualitative analysis.
AI-enhanced qualitative inquiry supports literature and systematic reviews, conceptualization, and thematic and content analysis (Christou, 2023, p. 1968). Large language models “are increasingly being used … to produce, translate, summarize, and analyze information” (Christou, 2023, p. 1969), automating tasks such as transcription, coding, and theme identification while enabling greater scale and precision (Hodges et al., 2025; Nicmanis & Spurrier, 2025; O’Connor et al., 2025). AI responds to longstanding challenges in qualitative work, where “research teams … may be overwhelmed with the volume of data and variabilities in interpretation” (O’Connor et al., 2025, p. 1).
Despite these gains, AI remains an assistive tool. Its role is still being defined (O’Connor et al., 2025, p. 2), and scholars caution that while AI can process large datasets efficiently, it has “limitations and weaknesses” (Christou, 2023, p. 1970). Emerging literature frames AI use as hybridization, where human interpretation and machine cognition complement one another (Olawade et al., 2025), p. 5). AI can streamline early analytic tasks, but it still relies on human reflexivity and contextual judgment; “even in analysis performed by AI … a researcher has an active role to play” ((Christou, 2023), p. 1976), as GPTs “may struggle with … nuanced information” (Christou, 2023), p. 1972).
The integration of AI into qualitative research initially emerged within healthcare and social science domains, deeply invested in understanding human experience, empathy, and communication. In healthcare, early efforts used AI to analyze patient narratives, clinician–patient interactions, and qualitative feedback to improve diagnostic decision-making and medical empathy (Al-Shoteri, 2022; Fazakarley et al., 2024; Henzler et al., 2025). Similarly, in the social sciences, researchers adopted AI and NLP to interpret large digital corpora such as social-media posts, online communities, and multimodal communication datasets (Fieldhouse et al., 2025; Ntsobi et al., 2024; Shahin, 2016).
1.1.2. Methodological Hybridization: Augmenting, Not Replacing, Interpretation
A defining feature of AI’s emergence in qualitative inquiry is methodological hybridization or the blending of algorithmic efficiency with human interpretive depth. Rather than supplanting qualitative traditions, AI serves as an analytical augmentor, supporting processes such as data organization, coding, and pattern recognition while leaving interpretive synthesis to human researchers (Perkins & Roe, 2024; Shahin, 2016).
Findings from O’Connor et al. (2025), p. 5) show that “AI-assisted analyses identified most themes accurately,” although “ChatGPT was less successful at locating subtle, interpretive themes but had equal success in reproducing concrete, descriptive themes.” Thus, AI enhances efficiency but requires researcher reflexivity to sustain contextual and theoretical depth. Olawade et al. (2025, p. 5) describe this as a “hybrid methodology that capitalizes on the complementary strengths of both traditional and digital methods while mitigating their individual limitations.”
1.1.3. Epistemic Tensions and Ethical Considerations
The adoption of AI in qualitative research has provoked reflection on epistemology, ethics, and methodological integrity. As Christou (2023, p. 1970) notes, “Researchers must acknowledge the strengths of AI … but also its limitations and weaknesses.” Human versus algorithmic reasoning remains a central tension: qualitative inquiry privileges situated interpretation and reflexivity, whereas machine learning relies on statistical regularities that may flatten context or nuance (Chatzichristos, 2025; Cruz-Aguilar, 2025; Williams, 2024). “AI can augment but not replace the work of human researchers in producing rigorous reflexive thematic analysis,” Hitch (Hitch, 2024, p. 597) emphasizes, while O’Connor et al. (2025, p. 2) conclude that “it is necessary for a human researcher to double check AI-completed work.”
1.1.4. The Expansion of AI-Assisted Qualitative Methods
AI has moved from experimental use to routine integration across qualitative workflows, shifting from isolated pilots to widespread methodological adoption. Studies show that NLP and neural models accelerate early-stage analysis—expediting coding, surfacing themes, and resolving code conflicts—while leaving interpretive synthesis to human researchers (Abdul Rahman et al., 2025; O’Connor et al., 2025).
In health research, AI-enabled sentiment analysis supports the interpretation of patient narratives by detecting emotional tone and subtle affective shifts (Olawade et al., 2025)), informing applications such as service redesign and empathy training (Abderrahim et al., 2024; Choi, 2020). AI is also used to combine qualitative insights with clinical data in decision-support systems, where explainable AI and multisource analytics streamline evidence synthesis and enable personalized care pathways (Pillay et al., 2025; Thomas & Kuppasani, 2025; Wen et al., 2025).
Across education and the social sciences, LLMs such as ChatGPT have become normalized analytical assistants, supporting literature review, summarization, focus-group simulation, and both deductive and inductive coding (Dengel et al., 2023; Hila & Hauser, 2025; Kondo et al., 2024; O’Connor et al., 2025; Wachinger et al., 2025; Yue et al., 2025). These tools can reveal patterns not immediately apparent to human analysts (Hitch, 2024).
1.1.5. Ethical Integrity, and Reflexive Human–AI Control
As AI becomes normalized in qualitative inquiry, concerns around reliability, validity, and interpretive authority have intensified. Although “ChatGPT produced summaries that were accurate and comparable to what researchers found,” human oversight remains indispensable because it is “less successful at locating subtle, interpretive themes” (O’Connor et al., 2025, p. 2). Scholars stress that interpretive control must remain with the researcher: AI outputs require carefully crafted prompts and must be “thoroughly reviewed by the analyst before any theoretical or conceptual discussion can take place” (Christou, 2023, p. 1977). Ensuring rigor therefore depends on transparent reporting of prompts, parameters, and workflows, alongside training to help researchers interpret AI-generated categorizations (Landerholm, 2025; O’Connor et al., 2025; Williams, 2024).
These concerns form part of a broader shift toward ethical governance in AI-assisted qualitative inquiry. Recent scholarship emphasizes that ethical integrity must be anchored in “transparency, accountability, reflexivity, and collaborative validation” (Costa et al., 2025, p. 9). Calls for comprehensive digital ethics protocols highlight risks such as opaque model provenance, gaps in consent for AI-mediated analysis, and a “normative vacuum” where technological development outpaces institutional regulation (Jobin et al., 2019; Olawade et al., 2025). Algorithmic bias—linguistic, cultural, and demographic—further threatens equity and validity, especially in healthcare and education where unequal error rates can reinforce disparities (Ismail & Ahmad, 2025; Lim et al., 2025; Olawade et al., 2025; Yan et al., 2025). Addressing these concerns requires interdisciplinary collaboration and stronger institutional policies (Cano, 2025; Olawade et al., 2025).
Emerging best-practice frameworks thus foreground informed consent, confidentiality, explainability, and robust data governance (Hitch, 2024; Lopez-Ramos et al., 2025). At the same time, scholars call for culturally sensitive and context-aware AI tools, particularly for low- and middle-income settings where linguistic and ethical considerations differ (Akuma et al., 2025; Fieldhouse et al., 2025; Olawade et al., 2025).
Collectively, these developments signal a transition from viewing AI as a mere tool toward conceiving it as a reflexive partner in meaning-making—one whose participation must remain transparent, auditable, and subordinate to human judgment. Reflexive models ask researchers to specify where AI enters the workflow, assess how it shapes interpretation, justify when AI suggestions are adopted or rejected, and document the ethical trade-offs at each step (Costa et al., 2025). Even as integration deepens, “a researcher has an active role to play” (Christou, 2023, p. 1976), providing the contextual, theoretical, and ethical reasoning that AI cannot supply.
1.2. Research Gap
Despite the rapid expansion of AI-assisted qualitative research, the literature remains fragmented, discipline-specific, and largely descriptive, offering rich accounts of tools, capabilities, and ethical concerns but little synthesis of how the field has evolved as a whole. Existing studies tend to examine AI through isolated case applications or conceptual reflections, leaving unanswered questions about the broader intellectual structure, cross-disciplinary convergence, and temporal shifts that define this emerging domain. Importantly, no study has systematically mapped how methodological, ethical, and pedagogical debates have coalesced into identifiable research themes—or how these themes have transformed from the early phase of experimentation to the current moment of ethical rebalancing and reflexive partnership. This gap underscores the need for a longitudinal science-mapping analysis that can capture not only what AI is doing in qualitative research, but how the field itself is reorganizing around these technologies. Addressing this gap informs both scholarly understanding and practical governance of AI in qualitative inquiry, thereby motivating the objectives that follow.
1.3. Research Questions
This study aims to systematically map the conceptual structure, thematic evolution, and epistemic trajectory of Artificial Intelligence (AI) in qualitative research from 2021 to 2025 using SciMAT. Specifically, it seeks answers to the following questions:
RQ1: What core themes and conceptual clusters define the landscape of AI in qualitative research across the periods 2021–2023, 2024, and 2025?
RQ2: How have these themes evolved over time and what do these shifts reveal about the field’s developmental trajectory?
RQ3: How do the evolving clusters reflect changes in the role of AI in qualitative inquiry?
RQ4: What methodological, ethical, and pedagogical gaps remain in the existing literature, and what future directions do the thematic patterns suggest?
2. Methodology
This scientometric study employed science mapping to analyze the thematic structure and longitudinal evolution of research (Bagheri et al., 2023; Cobo et al., 2011, 2012, 2018; Moral-Munoz et al., 2019) on AI in Qualitative Research. The analysis followed a co-word and thematic network approach to identify conceptual clusters, their interrelationships, and their evolution over time.
2.1. Data Retrieval and Screening
The dataset was sourced exclusively from the Scopus database, selected for its comprehensive coverage of peer-reviewed documents (Visser et al., 2021). Scopus has been validated in prior studies as a reliable repository for bibliometric analysis due to its broad coverage, quality control, and dependable metadata (Harzing & Alakangas, 2016; Zhu & Liu, 2020). The document retrieval was conducted last November 11, 2025. The following search string was used:
("artificial intelligence" OR "ai" OR "machine learning" OR "deep learning") AND ("qualitative research" OR "qualitative analysis" OR "qualitative methods" OR "qualitative data") AND ("data analysis" OR "thematic analysis" OR "content analysis" OR "coding")
The initial query returned 2,179 documents. To ensure temporal focus on the most recent developments in student engagement research in science education, the dataset was restricted to the 2021–2025 period, yielding 1,909 documents. After applying a language filter to include only English-language publications, 1,862 documents remained for analysis.
No filters were applied with respect to publication stage, document type, or source type to preserve the diversity of scholarly contributions, including journal articles, conference papers, book chapters, and reviews. The year-wise distribution of the dataset shows a marked upward trend: 106 for 2021, 174 for 2022, 240 for 2023, 528 for 2024, and 814 for 2025, reflecting a surging discourse on AI integration in qualitative research.
2.2. Descriptive Overview of the Dataset
The descriptive analysis (see Fig. 1) highlights distinct publication, disciplinary, and geographic trends in the emerging field of AI in qualitative research. The most active publication venues from 2021 to 2025 include Plos One, BMJ Open, Journal of Medical Internet Research, International Journal of Environmental Research and Public Health, and Lecture Notes in Networks and Systems. Among these, Journal of Medical Internet Research showed the most consistent growth, while Plos One experienced a sharp rise in 2024 followed by a slight decline in 2025, indicating shifting yet sustained interest in AI-assisted qualitative methods.
Fig. 1
Publication Trends and Disciplinary Distribution of Research on AI in Qualitative Research (2021–2025) (Source: Scopus)
In terms of geographic distribution, the United States leads with the highest number of publications, followed by China, the United Kingdom, India, and Canada. Other notable contributors include Australia, Germany, Turkey, Malaysia, and Saudi Arabia, reflecting the widespread adoption of AI methodologies in qualitative research across both Western and Asian regions.
Regarding disciplinary representation, the largest share of studies falls under Medicine (19.3%), followed closely by Social Sciences (17.3%), Computer Science (16.1%), and Engineering (7.3%). Additional contributions come from Business and Management (4.6%), Nursing (3.9%), Psychology (3.4%), Mathematics (3.2%), and Arts and Humanities (2.9%), demonstrating the interdisciplinary expansion of AI-assisted qualitative research across health, social, and computational sciences.
Institutional affiliations reveal that the University of Toronto and University College London are the most prolific contributors, followed by Monash University, the National University of Singapore, and the University of Toronto Faculty of Medicine. Other leading institutions include the University of Hong Kong, University of British Columbia, King’s College London, University of Queensland, and University of Manchester, illustrating strong engagement from major research universities across North America, Europe, Asia, and Oceania.
2.3. Analytical Tool and Workflow
The mapping and longitudinal analyses were conducted using SciMAT (Science Mapping Analysis Tool) (Cobo et al., 2012), following the standardized workflow illustrated in Fig. 2. The software integrates co-word analysis, strategic diagram construction, and thematic evolution mapping to identify conceptual structures and trace their development across defined periods.
Step 0: Word Grouping. Prior to analysis, keyword standardization was conducted using SciMAT’s built-in thesaurus functions—“Find Similar Words by Plurals” and “Find Similar Words by Distance.” A strict Levenshtein distance of 1 was applied to merge only near-identical terms, minimizing semantic distortion.
Step 1: Period Selection. The dataset was divided into three consecutive periods – period 1 covers the years 2021 to 2023 (n = 520), period 2 covers 2024 (n = 528), and period 3 covers 2025 (n = 814).
Fig. 2
Analytical Workflow of the SciMAT Co-Word and Longitudinal Mapping Process
Step 2: Unit of Analysis. The unit of analysis was set to words, specifically combining authors’ keywords and sources’ index keywords to represent both conceptual intention and indexing precision.
Step 3: Data Reduction. To focus on meaningful and recurrent themes, a minimum frequency threshold of three was applied. Only keywords appearing in at least three documents were retained for network construction.
Step 4: Kind of Matrix. A co-occurrence matrix was generated to represent conceptual linkages between keywords. This matrix forms the basis for visualizing how terms co-appear within documents, thereby revealing the conceptual architecture of the field.
Step 5: Network Reduction. To improve interpretability, weak connections were filtered out using an edge-value threshold of 3, retaining only the most significant conceptual ties.
Step 6: Normalization. Keyword co-occurrences were normalized using the Association Strength Index (Callon et al., 1983, 1991a, 1991b), which corrects for varying frequencies and ensures that relationships reflect meaningful conceptual proximity rather than raw frequency counts.
Step 7: Clustering Algorithm. Themes were identified using the Simple Centers Algorithm, which groups closely related keywords around core concepts to form thematic clusters. Each cluster represents a distinct thematic area within the research landscape.
Step 8: Document Mapper. Documents were automatically assigned to thematic clusters through SciMAT’s Core Mapper function, allowing for the identification of the most representative publications for each theme.
Step 9: Quality Measures. To evaluate the significance and maturity of each theme, bibliometric indicators such as the h-index and g-index were computed. These indicators provided a measure of both productivity and citation impact within each cluster.
Step 10: Longitudinal Analysis. The thematic evolution across periods was traced using the Association Strength measure for thematic continuity and the Jaccard’s Index for the overlapping map, which visualizes the degree of thematic stability, merging, or divergence over time.
Step 11: Thematic Analysis. Finally, the resulting strategic diagrams, evolution maps, and cluster networks were interpreted to identify motor themes, basic and transversal themes, emerging or declining themes, and niche themes.
The strategic diagrams plot themes along two dimensions: centrality, which measures the degree of interaction of a theme with other themes, and density, which measures the internal development and cohesion of the theme (Callon et al., 1991c). Based on their positions in the diagram, themes are classified into four groups:
Motor Themes (Q1): Highly developed and central, driving the field forward.
Basic and Transversal Themes (Q4): Essential and central but underdeveloped.
Specialized/Isolated Themes (Q2): Developed but marginal or isolated.
Emerging/Declining Themes (Q3): Weakly developed and peripheral, either nascent or fading.
2.4. Ethical Considerations
All data were retrieved from publicly accessible sources within Scopus. No personal or confidential information was involved, and analysis adhered to ethical research practices in scientometric studies, ensuring transparency, reproducibility, and accurate data citation.
3. Results
3.1. Period 1 (2021–2023): Thematic Structure and Intellectual Configuration
The thematic configuration of research on Artificial Intelligence (AI) in Qualitative Research during Period 1 (2021–2023) reflects an early yet decisive stage of convergence between traditional qualitative inquiry and computational approaches. The SciMAT strategic diagram reveals five major clusters—Semi-Structured Interview, Artificial Intelligence, Qualitative Research, Qualitative Analysis, and Human—each representing a distinct but interrelated conceptual domain.
A
Fig. 3
Strategic Diagram and Cluster Networks for Period 1 (2021–2023)
The Semi-Structured Interview cluster emerges as the period’s dominant motor theme, showing strong development and widespread connections. Its emphasis on machine-learning and NLP analysis of interview data—particularly in health research—demonstrates that traditional interviewing remains the methodological backbone even as AI assists with transcription, coding, and data organization. The Qualitative Research cluster, with moderate centrality, anchors the field’s epistemological core, reflecting continued reliance on applied qualitative methods and ongoing negotiation over how to integrate AI without compromising meaning, reflexivity, and contextual interpretation.
The Artificial Intelligence cluster is internally strong but conceptually peripheral, indicating a tool-centric rather than epistemic orientation in this early phase. AI functioned mainly as an auxiliary analytic aid, with limited integration into qualitative theory. Conversely, the Human cluster appears as an emerging but weakly developed theme, signaling early awareness of ethical and ontological tensions as AI mediates human experience, yet showing little conceptual consolidation. Qualitative Analysis sits in a transitional position, marking experimentation with AI-assisted analytic techniques that enhance, rather than replace, human interpretation and introduce early meta-analytic and bibliometric approaches.
Together, these clusters depict a field in early convergence: traditional qualitative practices remain dominant, computational tools enter cautiously, and tensions between human-centered interpretation and algorithmic rationality are only beginning to surface. Period 1 therefore represents an exploratory stage in which scholars test hybrid methodological possibilities, laying the groundwork for the more integrated and ethically mature configurations that develop in later periods.
3.2. Period 2 (2024): Thematic Consolidation and Emergent AI Integration
The SciMAT strategic diagram and cluster networks for Period 2 (2024) reveal a decisive phase of thematic consolidation and methodological transformation. Whereas Period 1 (2021–2023) was marked by exploratory hybridization, Period 2 shows a clear shift toward integration, diversification, and systematization. Clusters such as Interview, Patient Care, Human, Artificial Intelligence, Qualitative Analysis, Qualitative Research, and the new entrant ChatGPT demonstrate how the field is moving from the mere coexistence of human-centered and computational approaches toward a more coherent, practice-oriented ecosystem.
The Interview cluster becomes the dominant motor theme of 2024, expanding beyond semi-structured formats to encompass clinical communication and healthcare delivery. Its high centrality and density show that AI-supported interview analysis has become institutionalized, with tools now routinely aiding transcription, structuring, and preliminary coding. Closely linked is Patient Care, a mature applied theme where sentiment analysis, decision-support tools, and algorithmic text analysis enhance patient understanding and clinical empathy—though its concentration in health research highlights the need for broader theoretical development.
The Human cluster serves as a conceptual bridge across domains, reflecting growing ethical and interpretive concerns as scholars re-center human agency, emotion, and contextual meaning within AI-mediated workflows. Meanwhile, the Artificial Intelligence theme becomes more diversified, incorporating ethics, policy, and patient-centered care—marking a shift from purely technical applications toward questions of governance, responsibility, and value alignment.
Qualitative Analysis functions as the methodological pivot, showing strong internal cohesion and increasing hybridization of computational and interpretive approaches, including neural network–assisted patterning and emerging links to personalized medicine. By contrast, Qualitative Research remains an enduring epistemic anchor, though its lower density suggests theoretical discourse is struggling to keep pace with rapid AI-driven innovation. The distinct emergence of ChatGPT signals a transformative frontier, with LLMs now supporting transcript generation, code suggestion, conversational analysis, and early interpretive synthesis. Though still peripheral, its high density reflects rapid experimentation and raises new questions about authorship, reflexivity, and analytic rigor.
A
Fig. 4
Strategic Diagram and Cluster Networks for Period 2 (2024)
Collectively, Period 2 marks a shift from experimentation to systematized integration: AI becomes embedded across data collection, analysis, and epistemic construction, while human oversight and ethical reflexivity remain central. This thematic landscape reflects a maturing human–AI synergy in which meaning is co-produced through collaborative, rather than replacement-oriented, intelligence.
3.3. Period 3 (2025): Institutionalization, Human-Centric Ethics, and Pedagogical Expansion
The SciMAT strategic diagram and thematic cluster networks for Period 3 (2025) reveal a mature, interconnected, and ethically attuned research landscape for AI in Qualitative Research. Compared with the earlier exploratory (2021–2023) and consolidating (2024) phases, the 2025 period signifies a phase of epistemic stabilization—where human-centered, pedagogical, and ethical concerns are integrated into AI-mediated qualitative inquiry. The field has moved from experimenting with AI tools to institutionalizing frameworks that link technology with interpretive fidelity, professional practice, and reflective use.
A
Fig. 5
Strategic Diagram and Cluster Networks for Period 3 (2025)
The 2025 map is dominated by Health Personnel Attitude, now the central motor theme, indicating that AI-assisted qualitative analysis has become embedded in clinical practice. Its network—linking interviews, clinical guidelines, and EHRs—shows that AI is shaping clinicians’ decision-making, trust, and ethical reflection, marking a shift from methodological experimentation to everyday human–machine collaboration. Semi-Structured Interview continues as a core methodological anchor across all periods, now augmented by advanced AI tools that support coding, text mining, and emotional analysis while preserving human interpretive authority.
The Human cluster expands into a broader ethical and socio-contextual theme, foregrounding inclusivity, cultural bias, and global equity as central concerns. Its high centrality but low density indicates an evolving conceptual space where debates on oversight, justice, and lived experience converge. Qualitative Research likewise sustains the epistemic foundation of the field, increasingly linked to pedagogy and training as students and professionals learn to integrate AI literacy with qualitative reasoning.
The emergence of Students as a dense applied theme reflects the growing role of AI-assisted qualitative methods in education, highlighting experimentation with automated coding, AI-supported focus groups, and reflexive writing with LLMs. Qualitative Analysis becomes the technical backbone of hybrid inquiry, with deep neural networks and image-processing methods signaling porous boundaries between qualitative interpretation and computational modeling. Artificial Intelligence evolves into a reflexively governed theme emphasizing ethical standards, best practices, and responsible methodological integration. Meanwhile, ChatGPT remains an experimental frontier, used increasingly in pedagogical contexts and raising questions about interpretive agency and bias.
Synthesis: Thematic Evolution and Intellectual Trajectory of AI in Qualitative Research
The Evolution Map shows strong longitudinal continuity across core themes—Semi-Structured Interview, Qualitative Analysis, Qualitative Research, Human, and Artificial Intelligence—while new clusters such as Patient Care (2024), Health Personnel Attitude (2025), Students, and ChatGPT demonstrate diversification into applied, pedagogical, and generative AI domains. The Overlap Map confirms this expansion, with steady keyword retention (0.48 → 0.47) and growth in conceptual volume (474 → 552), signaling a field that is both coherent and rapidly evolving.
Phase 1 (2021–2023) marks the field’s experimental foundation. Semi-Structured Interview functioned as the methodological anchor that linked human-centered inquiry with early computational assistance. Qualitative Research provided epistemic grounding, while Artificial Intelligence remained technically advanced but conceptually peripheral. The Human cluster appeared fragmented, indicating early ethical and ontological tensions, and Qualitative Analysis served as a transitional site for the first attempts to hybridize interpretive methods with neural and image-based analytics. This period reflects an initial negotiation between algorithmic efficiency and human meaning-making.
Phase 2 (2024) shows clear systematization and applied integration. Interview became the new motor theme, reflecting the normalization of AI-augmented interviewing in clinical and organizational research. Patient Care emerged as a mature applied theme centered on AI-supported analysis of patient narratives and clinical communication. The Human cluster evolved into an integrative ethical framework, while Artificial Intelligence broadened toward policy, governance, and responsible innovation. The appearance of ChatGPT signaled the growing influence of LLMs in qualitative analysis and education. Qualitative Analysis and Qualitative Research remained stabilizing cores, balancing innovation with interpretive rigor.
A
Fig. 6
Evolution and Overlap Maps of AI in Qualitative Research (2021–2025)
Phase 3 (2025) represents institutional maturity and ethical rebalancing. Health Personnel Attitude became the dominant motor theme, indicating that AI-assisted qualitative methods are now embedded in professional practice and shaping clinicians’ judgments, trust, and emotional labor. Semi-Structured Interview persisted for a third period, underscoring methodological continuity. The Human cluster expanded ethically and globally, emphasizing inclusivity, cultural sensitivity, and social justice. Students and Qualitative Analysis highlighted the pedagogical and technical institutionalization of AI literacy, while Artificial Intelligence consolidated into a reflexive domain emphasizing best practices and ethical frameworks. ChatGPT remained an active frontier of experimentation within learning and qualitative reasoning.
Cross-Period Synthesis reveals a trajectory from convergence to coevolution. Stable thematic pathways (e.g., Semi-Structured Interview → Interview → Semi-Structured Interview; Artificial Intelligence → Artificial Intelligence across all periods) show preserved epistemic cores, while branching developments (Human → Students; Patient Care → Health Personnel Attitude) reflect adaptive diversification. Three meta-patterns characterize this evolution: (1) methodological continuity through adaptive hybridization, (2) a deepening human-centric reflexivity that safeguards ethics and context, and (3) pedagogical and professional institutionalization of AI-assisted qualitative inquiry.
4. Discussion
Across all three periods, interview-based designs and core qualitative constructs remain structurally central even as AI becomes more embedded in analytic workflows. This continuity aligns with Christou’s (2023) view that AI has entered qualitative inquiry mainly through thematic and content analysis, without displacing the interpretive frameworks that organize meaning-making. The persistent prominence of interview-focused and “qualitative research” themes suggests that AI is being adopted in ways that honor qualitative commitments to depth, context, and lived experience rather than yielding to purely algorithmic rationality. At the same time, the longitudinal expansion of AI-related clusters, from tool-oriented work to themes spanning clinical practice, patient care, and education, corroborates O’Connor et al.’s (2025) observation that AI is increasingly used to address bottlenecks of volume, time, and interpretive variability. The field’s evolution reflects the stabilization of “hybrid methodologies” (Olawade et al., 2025), in which AI supports transcription, coding, and pattern detection while interpretive synthesis remains under human control—a pragmatic rebalancing of labor rather than an epistemic surrender.
4.1. Methodological Hybridization and the Resilience of Interpretive Paradigms
Methodological hybridization has deepened without eroding the interpretive core of qualitative research. Early co-occurrences of qualitative analysis with neural networks and bibliometric techniques anticipate contemporary AI-assisted QDA tools that combine pattern recognition, clustering, and visualization with human coding, mirroring the transition from CAQDAS to AI-enhanced QDA (Ntsobi et al., 2024; Kikooma, 2010). The strategic prominence of qualitative analysis resonates with studies that treat AI as extending, not replacing, traditional logics: O’Connor et al. (2025) and Hamoud et al. (2025) show that AI-generated themes often align with human analyses for descriptive categories but falter with subtle or latent meanings. The centrality of hybrid clusters thus confirms at field level that innovation is occurring at the interface between AI’s capacity to scale and structure data and researchers’ ability to interpret nuance and contradiction. The continued presence of broad “qualitative research” themes, even as technical clusters grow, suggests that interpretivism, constructivism, and reflexivity remain key reference points, supporting Goldkuhl’s (2012) and Galuppo and Benozzo’s (2024) insistence that meaning is socially constructed and context-dependent.
4.2. From Ethical Anxiety to Reflexive Governance
The evolution of “Human” and “Artificial Intelligence” themes indicates how the field is grappling with ethics, bias, and governance. Early fragmentation of human-related themes reflects initial caution about authenticity, representation, and the risk of decontextualizing human experience in AI-mediated analysis (Cruz-Aguilar, 2025; Williams, 2024). Over time, these concerns consolidate into clusters foregrounding ethics, policy, and best practice, signaling a shift from diffuse anxiety to reflexive governance. This trajectory parallels Costa et al.’s (2025) call for transparency, accountability, and collaborative validation and Olawade et al.’s (2025) emphasis on algorithmic bias, linguistic and cultural inequities, and the need for digital ethics protocols tailored to qualitative work. The heightened visibility of themes related to health personnel attitudes, patient care, and global inequities suggests an expanding ethical aperture, echoing warnings that training-data bias and unequal error rates can entrench inequities (Yan et al., 2025; Ismail & Ahmad, 2025; Lim et al., 2025). AI increasingly appears not as neutral infrastructure but as a site where power, access, and voice must be continuously negotiated.
4.3. Pedagogical Turn: AI Literacy as a New Dimension of Qualitative Training
A distinctive contribution of this study is to highlight AI in qualitative research as a pedagogical issue. The consolidation of student-centered and educational clusters indicates that AI-assisted methods are becoming integral to how future researchers are trained, resonating with emerging work that frames AI as a scaffold for academic writing, literature scoping, and reflexive practice (Dengel et al., 2023; Wachinger et al., 2025). The coupling of student themes with ChatGPT and qualitative analysis suggests that LLMs are used to simulate focus groups, support coding exercises, and prompt reflexive writing, consistent with the affordances described by Christou (2023), Hitch (2024), and O’Connor et al. (2025). Yet the literature also warns that generic prompts yield generic outputs and that uncritical reliance on AI can erode interpretive rigor. The rise of pedagogical themes therefore underscores the need for explicit AI literacy: students must learn how to interrogate AI outputs, document AI involvement, and justify when to accept or reject AI-generated suggestions. As AI becomes woven into methods curricula, supervisors and educators will need frameworks that present AI as a co-analyst requiring reflexive monitoring rather than an invisible backstage assistant, even as theoretical discussions lag behind the pace of classroom experimentation.
4.4. Toward a Human-Centered AI–Qualitative Synergy
Overall, the longitudinal patterns suggest movement toward what Costa et al. (2025) term “AI as co-researcher,” albeit in a bounded sense. AI has shifted from peripheral tool to embedded infrastructure in clinical practice, research workflows, and education, but the persistent centrality of human-focused and qualitative clusters, along with expanding ethics and governance themes, points to a model where human judgment remains non-delegable. This aligns with Christou’s (2023) insistence that researchers retain responsibility for interpretation even when AI performs analytic tasks, and with Hitch’s (2024) argument that AI can augment but not replace reflexive thematic analysis. The field appears to be converging on an understanding of AI as a powerful yet constrained collaborator: it can scale, structure, and visualize data, but depends on human researchers for theoretical framing, contextual understanding, and ethical orientation.
4.5. Implications and Future Directions
Positioning the field’s thematic evolution within existing scholarship yields several implications. First, the resilience of interview-based and general qualitative themes suggests that future work should move beyond debating AI’s compatibility with qualitative inquiry toward specifying how particular tools align with distinct methodological traditions and epistemological commitments—for instance, comparing AI’s performance in narrative, phenomenological, or grounded-theory designs. Second, the prominence of ethics, human-centered themes, and global equity underscores the need for context-sensitive governance frameworks that address differentiated risks across settings, especially in low- and middle-income and multilingual contexts. Third, the emergence of pedagogical clusters implies that AI literacy should be treated as a core competence in qualitative education; more empirical work is needed on how students understand AI’s role in their analytic practice, how it affects agency and authorship, and what forms of guidance support critical use. Finally, the evolution of ChatGPT and related LLMs as distinct themes suggests that generative AI will remain a contested frontier, requiring robust criteria for when AI’s contributions constitute analytic labor, how they should be acknowledged, and where ethical boundaries lie.
The scientometric trajectory from 2021 to 2025 does not depict straightforward automation but an unfolding project of human-centered augmentation. AI has become integral to qualitative research infrastructures, yet its legitimacy depends on sustained efforts to ensure that interpretive depth, reflexivity, and justice remain at the heart of qualitative inquiry.
5. Conclusion
This study traced the intellectual evolution of AI-assisted qualitative research across three phases, revealing a field steadily reorganizing its methodological, ethical, and pedagogical foundations. The longitudinal maps demonstrate diffusion of new technologies and restructuring of qualitative inquiry itself. Over five years, researchers moved from tentative experimentation with AI-enhanced transcription and coding to systematized hybrid workflows and, ultimately, to institutional frameworks that embed ethics, pedagogy, and human-centered governance into AI-mediated analysis.
AI has expanded the operational capacity of qualitative traditions. Semi-structured interviews, narrative accounts, and interpretive reasoning remain the epistemic backbone of the field, even as NLP and large language models accelerate early-cycle analysis and reveal patterns at scale. Across periods, the evidence affirms that qualitative inquiry retains its identity when human researchers remain the architects of meaning-making.
A key contribution of this study is demonstrating how reflexive integration has become the methodological center of gravity. Researchers increasingly document where AI enters the workflow, how it shapes interpretation, and when algorithmic suggestions should be accepted or overridden. This transforms AI from a technical add-on into a transparent, auditable collaborator whose value depends on responsible use, ethical vigilance, and theoretical alignment.
The findings also reveal a widening pedagogical horizon. As AI literacy becomes foundational to research training, students learn to interrogate AI outputs, compare them with human interpretations, and cultivate ethical habits around disclosure, bias checking, and reflexivity. The emergence of Students and Health Personnel Attitude as major clusters indicates that AI-assisted qualitative inquiry now shapes how future clinicians, educators, and researchers learn to interpret human experience.
Finally, the strengthening of the Human cluster, particularly in global South and low-resource contexts, highlights the growing emphasis on contextual sensitivity and equity. The field increasingly recognizes that AI must adapt to linguistic, cultural, and socio-political diversity, not the other way around. Ensuring fairness, inclusivity, and contextual fit will be essential as AI deepens its role in qualitative work.
A
6. Declaration of AI UseOpenAI’s ChatGPT 5.1 was used in the preparation of this manuscript exclusively for language refinement, copy-editing, and structural improvement. The use of AI was limited to correcting grammar, enhancing clarity, and improving readability of the authors’ original text. No AI tools were employed to generate research ideas, analyze data, interpret findings, or draft substantive content beyond author-provided material. The author takes full responsibility for the accuracy, integrity, and originality of the work.
A
7. Conflict of Interest
The authors declare no known competing financial interests or personal relationships that could have appeared to influence the work reported in this study.
References
Abderrahim F, Mkik M, Ghernouk C, Khiati M, Aziz H, Hebaz A (2024) Revolutionizing Healthcare: The Impact of Artificial Intelligence in Connected Medicine–Unleashing the Power of Real-Time Diagnostics, Personalized Treatment and Ethical AI Adoption. In: Bansal JC, Borah S, Hussain S, Salhi S (eds) Lecture Notes in Networks and Systems: Vol. 1108 LNNS. Springer Science and Business Media Deutschland GmbH, pp 17–30. https://doi.org/10.1007/978-981-97-6588-1_2
Abdul Rahman H, Masri NA, Husaini A, Shaharuddin MY, Zaim SNN (2025) Integration of Machine Learning and Artificial Intelligence Techniques for Qualitative Research: The Rise of New Research Paradigms. International Journal of Qualitative Methods, 24. https://doi.org/10.1177/16094069251347545
Akuma I, Vaswani V, Ekmekci EP (2025) Ethical Implications of Artificial Intelligence in Vaccine Equity: Protocol for Exploring Vaccine Distribution Planning and Scheduling in Pandemics in Low- and Middle-Income Countries. JMIR Research Protocols, 14. https://doi.org/10.2196/76634
Al-Shoteri AH (2022) The Role of Methods and Applications of Artificial Intelligence Tools in the Field of Medicine to Diagnose and Discover Various Diseases. J Appl Data Sci 3(1):1–14. https://doi.org/10.47738/jads.v3i1.48
Bagheri B, Zadi H, Soltani A, Witlox F (2023) Global city data analysis using SciMAT: a bibliometric review. Environ Dev Sustain 26:15403–15427
Callon M, Courtial JP, Laville F (1991a) Co-word analysis as a tool for describing the network of interactions between basic and technological research: the case of polymer chemsitry. Scientometrics 22:155–205
Callon M, Courtial JP, Laville F (1991b) Co-word analysis as a tool for describing the network of interactions between basic and technological research: The case of polymer chemsitry. Scientometrics 22(1):155–205. https://doi.org/10.1007/BF02019280
Callon M, Courtial JP, Laville F (1991c) Co-word analysis as a tool for describing the network of interactions between basic and technological research: The case of polymer chemsitry. Scientometrics 22(1):155–205. https://doi.org/10.1007/BF02019280
Callon M, Courtial J-P, Turner WA, Bauin S (1983) From translations to problematic networks: An introduction to co-word analysis. Social Sci Inform 22(2):191–235. https://doi.org/10.1177/053901883022002003
Cano CAG (2025) Research, Ethics and Artificial Intelligence Challenges and Opportunities. In: Farhaoui Y, Herawan T, Imoize AL, Allaoui AE (eds) Lecture Notes in Networks and Systems: Vol. 1353 LNNS. Springer Science and Business, pp 487–497. https://doi.org/10.1007/978-3-031-88304-0_68Media Deutschland GmbH
Chatzichristos G (2025) Qualitative Research in the Era of AI: A Return to Positivism or a New Paradigm? International Journal of Qualitative Methods, 24. https://doi.org/10.1177/16094069251337583
Choi K-S (2020) Integrating artificial intelligence into healthcare research. J Nurs 67(5):12–18. https://doi.org/10.6224/JN.202010_67(5
Christou P (2023) Ηow to Use Artificial Intelligence (AI) as a Resource, Methodological and Analysis Tool in Qualitative Research? The Qualitative Report. https://doi.org/10.46743/2160-3715/2023.6406
Cobo MJ, Jürgens B, Herrero-Solana V, Martínez MA, Herrera-Viedma E (2018) Industry 4.0: A perspective based on bibliometric analysis. Procedia Comput Sci 139:364–371. https://doi.org/10.1016/j.procs.2018.10.278
Cobo MJ, López-Herrera AG, Herrera-Viedma E, Herrera F (2011) An approach for detecting, quantifying, and visualizing the evolution of a research field: A practical application to the Fuzzy Sets Theory field. J Informetrics 5(1):146–166. https://doi.org/10.1016/j.joi.2010.10.002
Cobo MJ, Lõpez-Herrera AG, Herrera-Viedma E, Herrera F (2012) SciMAT: A new science mapping analysis software tool. J Am Soc Inform Sci Technol 63(8):1609–1630. https://doi.org/10.1002/asi.22688
Costa AP, Bryda G, Christou PA, Kasperiuniene J (2025) AI as a Co-researcher in the Qualitative Research Workflow: Transforming Human-AI Collaboration. International Journal of Qualitative Methods, 24. https://doi.org/10.1177/16094069251383739
Cruz-Aguilar MA (2025) The epistemic revolution of AI: reconfiguring the foundations of scientific knowledge. AI Soc. https://doi.org/10.1007/s00146-025-02658-3
Dengel A, Gehrlein R, Fernes D, Görlich S, Maurer J, Pham HH, Großmann G, Eisermann NDG (2023) Qualitative Research Methods for Large Language Models: Conducting Semi-Structured Interviews with ChatGPT and BARD on Computer Science Education. Informatics 10(4). https://doi.org/10.3390/informatics10040078
Dodgson JE (2019) Reflexivity in Qualitative Research. J Hum Lactation 35(2):220–222. https://doi.org/10.1177/0890334419830990
Fazakarley C-A, Breen M, Thompson B, Leeson P, Williamson V (2024) Beliefs, experiences and concerns of using artificial intelligence in healthcare: A qualitative synthesis. Digital Health, 10. https://doi.org/10.1177/20552076241230075
Fieldhouse JK, Ge J, Randhawa N, Wolking D, Genovese BN, Mazet JAK, Desai A (2025) The intersection of artificial intelligence with qualitative or mixed methods for communicable disease research: a scoping review. Public Health, 248. https://doi.org/10.1016/j.puhe.2025.105961
Galuppo L, Benozzo A (2024) Qualitative research. Elgar Encyclopedia of Organizational Psychology. Edward Elgar Publishing Ltd, pp 573–578. https://doi.org/10.4337/9781803921761.00113
Gillan C, Palmer C, Bolderston A (2014) Qualitative methodologies and analysis. In Research for the Radiation Therapist: From Question to Culture (pp. 127–152). Apple Academic Press. https://www.scopus.com/inward/record.uri?eid=2-s2.0-85054208035&partnerID=40&md5=adc14e465f46c26e67f99db85e02bd70
Goldkuhl G (2012) Pragmatism vs interpretivism in qualitative information systems research. Eur J Inform Syst 21(2):135–146. https://doi.org/10.1057/ejis.2011.54
Harzing A-W, Alakangas S (2016) Google Scholar, Scopus and the Web of Science: a longitudinal and cross-disciplinary comparison. Scientometrics 106(2):787–804. https://doi.org/10.1007/s11192-015-1798-9
Henzler D, Schmidt S, Koçar A, Herdegen S, Lindinger GL, Maris MT, Bak MAR, Willems DL, Tan HL, Lauerer M, Nagel E, Hindricks G, Dagres N, Konopka MJ (2025) Healthcare professionals’ perspectives on artificial intelligence in patient care: a systematic review of hindering and facilitating factors on different levels. BMC Health Serv Res 25(1). https://doi.org/10.1186/s12913-025-12664-2
Hila A, Hauser E (2025) Assessing the Reliability of Large Language Models for Deductive Qualitative Coding: A Comparative Intervention Study with ChatGPT. Proceedings of the Association for Information Science and Technology, 62(1), 275–285. https://doi.org/10.1002/pra2.1255
Hitch D (2024) Artificial Intelligence Augmented Qualitative Analysis: The Way of the Future? Qual Health Res 34(7):595–606. https://doi.org/10.1177/10497323231217392
Hodges A, Ponce T, Seawright L (2025) AI-Assisted Qualitative Coding: Comparing Human to Machine Outputs. IEEE International Professional Communication Conference, 230–233. https://doi.org/10.1109/ProComm64814.2025.00051
Ismail O, Ahmad N (2025) Ethical and Governance Frameworks for Artificial Intelligence: A Systematic Literature Review. Int J Interact Mob Technol 19(14):121–136. https://doi.org/10.3991/ijim.v19i14.56981
Jobin A, Ienca M, Vayena E (2019) The global landscape of AI ethics guidelines. Nat Mach Intell 1(9):389–399. https://doi.org/10.1038/s42256-019-0088-2
Kikooma JF (2010) Using qualitative data analysis software in a social constructionist study of entrepreneurship. Qualitative Res J 10(1):40–51. https://doi.org/10.3316/QRJ1001040
Kondo T, Miyachi J, Jönsson A, Nishigori H (2024) A mixed-methods study comparing human-led and ChatGPT-driven qualitative analysis in medical education research. Nagoya J Med Sci 86(4):620–644. https://doi.org/10.18999/nagjms.86.4.620
Landerholm A (2025) AI in Qualitative Health Research Appraisal: Comparative Study. JMIR Formative Research, 9. https://doi.org/10.2196/72815
Liamputtg P (2009) Qualitative data analysis: Conceptual and practical considerations. Health Promotion J Australia 20(2):133–139. https://doi.org/10.1071/he09133
Lim T, Gottipati S, Cheong M (2025) What students really think: unpacking AI ethics in educational assessments through a triadic framework. Int J Educational Technol High Educ 22(1). https://doi.org/10.1186/s41239-025-00556-8
Lopez-Ramos LM, Pluktaite G, Bui CKT, Amann J, Haven T, Madai VI (2025) Operationalizing AI ethics in medicine-a co-creation workshop study. BMC Med Ethics 26(1):150. https://doi.org/10.1186/s12910-025-01317-y
Moral-Munoz JA, López-Herrera AG, Herrera-Viedma E, Cobo MJ (2019) Science mapping analysis software tools: A review. In Springer Handbooks (pp. 159–185). Springer. https://doi.org/10.1007/978-3-030-02511-3_7
Namatovu R, Langevang T (2025) Reflexivity in qualitative research engaging with young people in post-war northern Uganda. Handbook for Qualitative Research in Emerging Markets: Methods and Applications. Edward Elgar Publishing Ltd, pp 236–249. https://doi.org/10.4337/9781035322329.00026
Nicmanis M, Spurrier H (2025) Getting Started with Artificial Intelligence Assisted Qualitative Analysis: An Introductory Guide to Qualitative Research Approaches with Exploratory Examples from Reflexive Content Analysis. International Journal of Qualitative Methods, 24. https://doi.org/10.1177/16094069251354863
Ntsobi MP, Costa AP, Kasperiuniene J, Brandão C, Ribeiro J (2024) Digital Tools and Techniques in Qualitative Research: Digital Skills and Research Optimisation. In: Ribeiro J, Brandão C, Ntsobi M, Kasperiuniene J, Costa AP (eds) Lecture Notes in Networks and Systems: Vol. 1061 LNNS. Springer Science and Business Media Deutschland GmbH, pp 1–25. https://doi.org/10.1007/978-3-031-65735-1_1
O’Connor SK, Miller EE, Zweifel AR, Schievelbein DM, Parmar AR, Amell JW (2025) Use of artificial intelligence processing tools to evaluate qualitative data: Student researchers compared to faculty researchers. Currents Pharm Teach Learn 17(10):102418. https://doi.org/10.1016/j.cptl.2025.102418
Olawade DB, Omeni D, Gore MN, Hadi M (2025) Enhancing qualitative research through virtual focus groups and artificial intelligence: A review. Int J Med Informatics 203:106004. https://doi.org/10.1016/j.ijmedinf.2025.106004
Olmos-Vega FM, Stalmeijer RE, Varpio L, Kahlke R (2023) A practical guide to reflexivity in qualitative research: AMEE Guide 149. Med Teach 45(3):241–251. https://doi.org/10.1080/0142159X.2022.2057287
Peake L, Koleth E (2024) Feminist approaches to qualitative data analysis. In Doing Feminist Urban Research: Insights from the GenUrb Project. Taylor Francis 257–272. https://doi.org/10.4324/9781032668727-22
Perkins M, Roe J (2024) The use of Generative AI in qualitative analysis: Inductive thematic analysis with ChatGPT. J Appl Learn Teach 7(1). https://doi.org/10.37074/jalt.2024.7.1.22
Pillay TS, Topcu Dİ, Yenice S (2025) Harnessing AI for enhanced evidence-based laboratory medicine (EBLM). Clinica Chimica Acta, 569. https://doi.org/10.1016/j.cca.2025.120181
Pope C, Ziebland S, Mays N (2019) Analysis. In Qualitative Research in Health Care (pp. 111–133). wiley. https://doi.org/10.1002/9781119410867.ch9
Sarfo JO, Attigah DK (2025) Reflecting on Reflexivity and Positionality in Qualitative Research: What, Why, When, and How? J Advocacy Res Educ 12(1):75–81. https://doi.org/10.13187/jare.2025.1.75
Shahin S (2016) When Scale Meets Depth: Integrating Natural Language Processing and Textual Analysis for Studying Digital Corpora. Communication Methods Measures 10(1):28–50. https://doi.org/10.1080/19312458.2015.1118447
Sirris S (2022) Researchers’ role reflexivity when studying values work. In Researching Values: Methodological Approaches for Understanding Values Work in Organisations and Leadership (pp. 205–224). Springer International Publishing. https://doi.org/10.1007/978-3-030-90769-3_12
Thomas A, Kuppasani PR (2025) Advantages and Disadvantages of Artificial Intelligence. In Artificial Intelligence and Biological Sciences (pp. 93–110). CRC Press. https://www.scopus.com/inward/record.uri?eid=2-s2.0-105002513005&partnerID=40&md5=f357fdb97db189b85ccb0cd88fda4a42
Thompson D, Deatrick JA, Knafl KA, Swallow VM, Wu YP (2022) A Pragmatic Guide to Qualitative Analysis for Pediatric Researchers. J Pediatr Psychol 47(9):1019–1030. https://doi.org/10.1093/jpepsy/jsac040
Visser M, Eck NJ, Waltman L (2021) Large-scale comparison of bibliographic data sources: Scopus, web of science, dimensions, crossref, and microsoft academic. Quant Sci Stud 2(1):20–41
Wachinger J, Bärnighausen K, Schäfer LN, Scott K, McMahon SA (2025) Prompts, Pearls, Imperfections: Comparing ChatGPT and a Human Researcher in Qualitative Data Analysis. Qual Health Res 35(9):951–966. https://doi.org/10.1177/10497323241244669
Wen B, Wu Y, Daqqaq T, Chaddad A (2025) Towards a transparent and interpretable AI model for medical image classifications. Cogn Neurodyn 19(1). https://doi.org/10.1007/s11571-025-10343-w
Whitley R (2008) Social capital and public health: Qualitative and ethnographic approaches. In Social Capital and Health (pp. 95–115). Springer New York. https://doi.org/10.1007/978-0-387-71311-3_6
Williams RT (2024) Paradigm shifts: exploring AI’s influence on qualitative inquiry and analysis. Frontiers in Research Metrics and Analytics, 9. https://doi.org/10.3389/frma.2024.1331589
Williamson K, Given LM, Scifleet P (2018) Qualitative data analysis. In Research Methods: Information, Systems, and Contexts: Second Edition (pp. 453–476). Elsevier Inc. https://doi.org/10.1016/B978-0-08-102220-7.00019-4
Yan Y, Liu H, Chau T (2025) A Systematic Review of AI Ethics in Education: Challenges, Policy Gaps, and Future Directions. J Global Inform Manage 33(1). https://doi.org/10.4018/JGIM.386381
Yue Y, Liu D, Lv Y, Hao J, Cui P (2025) A Practical Guide and Assessment on Using ChatGPT to Conduct Grounded Theory: Tutorial. J Med Internet Res 27(1). https://doi.org/10.2196/70122
Zhu J, Liu W (2020) A tale of two databases: the use of Web of Science and Scopus in academic papers. Scientometrics 123(1):321–335