Sage Journals: Discover world-class research

Abstract

Background

The exponential growth of artificial intelligence (AI) in healthcare has raised critical ethical concerns.

Aims

To systematically identify research hotspots and trends of AI ethics in the medical field and provide evidence-based insights for future research.

Methods

A bibliometric analysis was conducted on publications from the Web of Science Core Collection (WOSCC) and China National Knowledge Infrastructure (CNKI) up to August 24, 2025. Visualisation tools were used to map the publication trend, as well as author, country, institution, journal, and keyword distributions. Keyword co-occurrence networks, clustering and burst analysis were employed to identify research hotspots and evolving trends. The results are reported following the BIBLIO checklist.

Results

A total of 1034 publications (291 from CNKI, 743 from WOSCC) were included. Key journals were Chinese Medical Ethics (CNKI) and Journal of Medical Ethics (WOSCC). China (n = 291) and the USA (n = 174) were the top publishing countries in CNKI and WOSCC, respectively. Leading institutional output came from developed countries (e.g., Harvard University, University of London, University of Toronto). Four research hotspots were identified: 1) ethical issues in different application scenarios of AI techniques in clinical practice, 2) AI ethical concerns and considerations in medical academia, 3) ethical challenges in AI-driven medical education, and 4) ethical governance, supervision, and review in healthcare. Keyword burst analysis indicated an increasing attention on balancing AI development with ethical governance, promoting research on complex intelligent tools embedded with ethical principles.

Conclusion

Attention to this topic has grown steadily over the past eight years. Research trends reflect a shift from fragmentation to the integration of technology and ethics in medicine. Future research should further refine the operational definitions of ethical principles across diverse application scenarios to guide stakeholders in effectively integrating technology and ethics.

Keywords

Artificial intelligence medicine ethics CiteSpace visual analysis

Introduction

The rapid advancement of artificial intelligence (AI) is driving a profound transformation in healthcare. AI technology has permeated various critical domains, including disease prediction, screening, diagnosis, treatment, and surveillance, medical imaging analysis, predictive analytics and risk assessment, clinical decision support, robot-related research, virtual and remote assistance, mental health support, medical education, training and consultation.^1–5 These applications enhance diagnostic accuracy, improve work efficiency, and ultimately lead to better patient outcomes.¹ However, this proliferation has also raised pressing ethical concerns, including challenges to the integrity of informed consent,⁶ data security and privacy risks, algorithmic bias and opacity, rigid regulatory frameworks, insufficient rigour in ethical review processes, gaps in governance, and the issues of patient acceptance and trust in AI solutions, all of which demand immediate and thoughtful resolution.^7,8

These pressing challenges have garnered significant academic attention. Scholars have conducted bibliometric analyses on this topic from diverse perspectives, covering health big data privacy protection,⁹ ethical issue categorisation in AI-driven healthcare,¹⁰ AI advancement in health and medicine,^11,12 AI trends in critical care medicine¹³ and AI-enabled medical education.^14,15 Overall, from a macroscopic perspective, the focus of these studies is relatively dispersed, hindering a comprehensive grasp of this field. Furthermore, most studies tend to describe research hotspots from a static perspective, with no trend analysis of hot keywords. Additionally, most review papers exhibit language bias due to their reliance on monolingual sources.^16–18 This often overlooks valuable research published in other languages (e.g., Chinese), creating a gap in the literature.

Therefore, this study aims to address the aforementioned research gaps by incorporating publications from both English and Chinese academic databases. This approach leverages the complementary nature of these sources to mitigate language bias and provide a more comprehensive and systematic overview of dynamic global research hotspots and evolving trends. The findings are expected to offer valuable insights for future research, standardisation, and policy formulation. This will help safeguard patient rights and public interests, ensuring that medical AI technology contributes to the well-being of human society.

Materials and methods

Search strategy

The China National Knowledge Infrastructure (CNKI) and Web of Science Core Collection (WOSCC) databases were systematically searched from database inception to August 24, 2025. The search strategy was developed through discussions among the authors (JLL and BLL), based on a review of similar literature and expert opinions (QSZ) in the field of evidence-based medicine. Search terms encompassed “AI”, “artificial intelligence”, “medic”, “medics”, “medical”, “medicine”, “ethic”, and “ethics”. Considering that the “ethic” and “medic” have multiple lexical variants, truncation symbols were incorporated in our search query (WOSCC). The detailed search strategies are provided in Supplementary File 1.

Eligibility criteria

Two authors (JLL and BLL) established the selection criteria through deliberation. We aimed to include studies on the ethics of AI in clinical settings. The detailed eligibility criteria are presented in Table 1.

Table 1.

Eligibility criteria.

Database	Inclusion criteria	Exclusion criteria
WOSCC	Topic: ethics of AI in healthcare Time: from the database inception to August 24, 2025 Language: English or Chinese Document types: articles or review articles	Duplicate publications
CNKI	Topic: ethics of AI in healthcare Time: from the database inception to August 24, 2025 Language: English or Chinese Document types: journal articles (original research or reviews)	Duplicate publications

Study screening and selection

Two researchers (JLL and BLL) independently performed the initial search, yielding 1403 records from WOSCC and 448 records from CNKI. Following the eligibility criteria, a total of 1537 records (364 from CNKI, 1173 from WOSCC) were imported into EndNote 20.0 for screening. After removing duplicates, further screening was conducted by JLL and BLL, who comprehensively reviewed titles, abstracts, and keywords. Discrepancies were resolved through consensus discussions within the research team to confirm thematic relevance (JLL, BLL, and YJQ). This multistage selection process resulted in 291 Chinese and 743 English publications (1034 total) for subsequent analysis. The entire process of study searching and selection is shown in Figure 1. Related records are shown in Supplementary File 2.

Figure 1.

Studies screening and selection flow diagram.

Data extraction and import

The titles, abstracts, authors, keywords, and other metadata of eligible publications were extracted and exported. Chinese publications were formatted in RefWorks, while English publications were in RIS format. Both datasets were saved as “download_.txt” files and imported into CiteSpace 6.4.R1 software for analysis. WOSCC data were directly compatible with the software, whereas CNKI data required format conversion to ensure interoperability. Given the language difference in data format and structure between WOSCC and CNKI, separate data import and analysis were conducted for each database.

Data analysis

Data analysis included descriptive analysis and visual analysis. Data were initially collated and summarised in Microsoft Excel for descriptive analysis, which included the annual trends in publication output and the distribution of the number of journal publications. Author, country, institution, and keyword (co-occurrence, clustering, and burst) visualisation analyses were primarily performed using CiteSpace software. A 1-year time slice was adopted, with the time spans set as 1999–2025 for WOSCC and 2017–2025 for CNKI. Term extraction sources included title, abstract, author keywords (DE), and keywords plus (ID). Nodes were sequentially selected as author, institution, country, and keyword for corresponding analysis. The g-index threshold was set to k = 25, while all other parameters were configured as default values. Detailed software configurations can be found in Supplementary File 3.

Results

Distribution of publication time and journals

Annual publication trends

Publications in this field have demonstrated a gradual upward trend overall. The earliest record in CNKI dates back to 2017, focusing on AI applications in biomedicine and strategies for addressing legal and ethical challenges.¹⁹ In contrast, the earliest record in WOSCC dates back to 1999, with initial studies concentrating on the ethical implications of electronic health records.²⁰ Following these early publications, both databases experienced a prolonged period of low output. Sustained growth began around 2017, with a more pronounced increase after 2018. In comparison, the growth within CNKI has been more gradual. Given that data for 2025 are incomplete (cutoff date: 24 August 2025), we tentatively identify 2025 as the peak year for Chinese publications (Figure 2).

Figure 2.

Annual trends in publication output.

Ranking of the number of journal publications

Journals with the highest publication output in this field are presented in Table 2. For CNKI publications, Chinese Medical Ethics leads with 38 articles, followed by Medicine and Philosophy. For WOSCC publications, the Journal of Medical Ethics has the highest output (37 articles), followed by the Journal of Medical Internet Research.

Table 2.

Key journals and publication counts.

Journals (CNKI)	Number of published articles	Journals (WOSCC)	Number of published articles
Chinese Medical Ethics	38	J Med Ethics (UK)	37
Medicine and Philosophy	36	J Med Internet Res (Canada)	22
Medicine and Society	3	BMC Med Ethics (UK)	17
Journal of Dialectics of Nature	2	Healthcare (Switzerland)	16
China Medical Equipment	2	BMC Med Educ (UK)	15

Publication volume of country/institution

In WOSCC, the top five countries by publication count were the USA (174 articles), Germany (62 articles), China (55 articles), the UK (50 articles), and Australia (40 articles). Since all publications in CNKI (291) were from China, a cross-database deduplication was performed. The results confirm that China and the US are the top publishing countries in CNKI and WOSCC, respectively. Research on the ethical issues of AI in the medical field was initiated earlier in developed countries. However, the substantial publication volume from China indicates a growing focus on this topic among major developing economies.

Further analysis of institutional publication output shows that Harvard University (USA) and University College London (UK) rank first and second, with 25 and 24 articles published, respectively, followed by the University of Toronto (Canada) in third place with 22 articles. Notably, although China leads globally in total publications, no Chinese institution ranks among the top global publishers. Within China, the most prolific institution is the School of Humanities and Social Sciences at Peking Union Medical College, with four articles (Table 3).

Table 3.

Ranking of the number of publications.

Research institutions (CNKI)	Number of publications	Research institutions (WOSCC)	Number of publications
School of Humanities and Social Sciences, Peking Union Medical College	4	Harvard University (USA)	25
China National Centre for Biotechnology Development	3	University of London (UK)	24
School of Marxism, Shanghai Jiao Tong University	2	University of Toronto (Canada)	22
School of Medical Humanities, Peking University	2	University of California System (USA)	18
Peking Union Hospital, Chinese Academy of Medical Sciences & Peking Union Medical College	2	University of Oxford (UK)	16

Node type: institution.

Research team and author analysis

In the WOSCC database, prominent research teams include those led by Buyx (Technical University of Munich, Germany) and Savulescu (National University of Singapore, Singapore), each with seven publications. Buyx's team focuses on embedding ethics into the development of AI in medicine, whereas Savulescu's team focuses on ethical considerations pertaining to AI applications in medical settings.

In CNKI, the leading research team is headed by Fei Xu from the University of Science and Technology of China, having published four articles to date. Their research primarily addresses gaps in the soundness, maturity, supervision, and governance of ethical norms, as well as innovations in embedding ethical values, formulating ethical practice guidelines, and improving review mechanisms. A co-authorship analysis indicates that Xin Huang and Tianwei Yuan, who rank among the most productive, have collaborated on ethical issues in medical safety, life sciences, biology, ecological safety, and other domains in the context of emerging medical technologies (including AI, gene editing, 3D printing, and biotechnology).

Frequency of hot keywords

Analysis of WOSCC data shows that the top five most frequent keywords are “AI”, “medical ethics”, “machine learning”, “medical informatics”, and “general & internal medicine”. Analysis of CNKI data reveals the top keywords to be “AI”, “medical ethics”, “ethics”, “ethical review”, and “ethical issues”. These findings indicate that scholars pay substantial attention to ethical issues of AI in clinical application (WOSCC) and the construction of ethical governance, review, and regulatory frameworks (CNKI). Detailed keyword co-occurrence patterns are presented in Figure 3(a) and (b), respectively.

Figure 3.

Co-occurrence of hot keywords (a) co-occurrence of hot keywords in CNKI, Node = 153, Link = 297; and (b) co-occurrence of hot keywords in WOSCC, Node = 392, Link = 1557. Node type: keyword, Link: connection of keywords.

Keyword co-occurrence analysis

Keyword co-occurrence analysis was performed on the 291 CNKI and 743 WOSCC publications, resulting in networks of 153 nodes and 297 connections for CNKI, and 392 nodes and 1557 links for WOSCC. In these networks, nodes represent keywords and links represent co-occurrence relationships. Key metrics include “frequency” (the occurrence count of a keyword), “betweenness centrality” (the positional importance of a node in the network), and “year” (the year the keyword first appeared).²¹ Key nodes were selected based on high frequency and a betweenness centrality value of ≥0.1. Notably, some nodes in the CNKI network had betweenness centrality values exceeding 1. As defined by Chen,²² betweenness centrality measures the proportion of all shortest paths in a network that pass through a given node. Following verification, we applied max normalisation for normalisation (Table 4). The detailed raw and normalised values are shown in Supplementary File 4.

Table 4.

Statistics of keywords with high centrality rankings.

Keywords (CNKI)	Frequency	Centrality (normalisation)	Year	Keywords (WOSCC)	Frequency	Centrality (normalisation)	Year
Artificial intelligence	86	1	2018	Artificial intelligence	442	1	2017
Medical ethics	22	0.14	2019	Computer science	91	0.26	1999
Ethics (norms)	14	0.09	2019	Machine learning	114	0.24	2019
Ethical issues	13	0.08	2017	Medical ethics	130	0.22	2019
Ethical review	11	0.06	2019	Medical informatics	109	0.22	1999
Ethical principles	7	0.06	2020	Big data	58	0.22	2019
Ethical governance	7	0.05	2019	General & internal medicine	93	0.15	2019
Intelligent medicine	7	0.04	2021	Education & educational research	57	0.14	2021
Supervision	2	0.04	2019	Biomedical social sciences	66	0.13	2020
Ethical risk	5	0.03	2020	Digital health	30	0.13	2019
Smart medicine	5	0.03	2018	Deep learning	35	0.11	2020
Ethical Dilemma	4	0.03	2022	AI ethics	35	0.10	2019
Doctor-patient relationship	4	0.02	2019	Large language models	21	0.10	2023
Justice	3	0.02	2020	Medical education	34	0.07	2023
Dignity	3	0.02	2022	Environmental & occupational health	30	0.07	2020
Fairness	2	0.02	2020	Informed consent	27	0.07	2021
Talent cultivation	2	0.02	2019	Generative AI	11	0.06	2024
Ethics (discipline)	2	0.02	2020	Medical AI	19	0.04	2022
Medical Humanities	2	0.02	2020	Precision medicine	13	0.04	2019
				Explainable AI	8	0.04	2024
				Genetics & heredity	5	0.04	2019
				Social issues	50	0.03	2021
				Biomedical social	22	0.03	2020
				Medical services	15	0.03	2023
				Decision making	15	0.03	2023
				Medical students	12	0.03	2023
				Health care	8	0.03	2023
				Clinical decision support systems	7	0.03	2021
				Health data	6	0.03	2023
				Biotechnology & applied microbiology	3	0.03	2019

Node type: keyword. Normalisation formula: X_normalised = (x − x_min)/(x_max − x_min), X∈[0, 1].

For WOSCC studies, key research hotspots include: 1) Ethical challenges associated with the application of core AI technologies (e.g., machine learning, big data, deep learning, large language models) in healthcare settings; 2) AI applications within specific healthcare subfields; 3) Core research focuses and challenges (e.g., informed consent, social issues, health data); 4) Ethical dimensions of AI integration in medical education.

For CNKI studies, key research hotspots centre on: 1) Ethical issues, dilemmas, and risks of AI applications in healthcare; 2) Ethical review frameworks, monitoring, and governance for intelligent healthcare systems; 3) Ethical issues in talent cultivation, medical humanities, and medical education; 4) AI-driven transformations in physician-patient relationships and principles of justice, dignity, and fairness in medicine.

In summary, there is a global focus on the ethical challenges of AI in clinical practice and medical education. We further reviewed articles associated with key nodes (from CNKI and WOSCC) and highly cited papers identified through reference analysis in WOSCC (Supplementary File 5). A core topic common to both databases is AI-driven academic issues in medicine. Methodologically, studies in the CNKI database primarily employed literature analysis or descriptive reviews to discuss the construction of ethical review and governance. In contrast, studies in the WOSCC database employed a wider variety of research methodologies to explore a broader range of topics (Table 5).

Table 5.

Diverse methodologies in WOSCC.

Research type	Topics
Qualitative studies	Ethical perspectives of academic machine learning researchers on healthcare algorithm development²³Value conflicts among stakeholders on ethical challenges²⁴Impacts of AI on the physician–patient relationship from stakeholders’ perspectives²⁵Patient perceptions of AI in caregiving relationships and ethics²⁶Ethical problem-solving through structured stakeholder dialogue²⁷Exploration of an ethical model for guiding clinical implementation²⁸
Cross-section studies	Assessment of stakeholders’ perception²⁹, understanding³⁰, knowledge and attitudes³¹, and acceptability³² of AI and ethicsDevelopment and validation of relevant scales³³
Mixed-method studies	Views of medical doctors on the ethics of AI³⁴Potential utility of AI large language models in medical ethics³⁵
A case study	Detection of ethical concerns in medical research leveraging AI³⁶
Observational study	Diagnostic accuracy of AI in an emergency room setting³⁷
A position paper	Ethical analysis and commentary on AI-related medical trials³⁸
Review	Ethical considerations in randomised controlled trials of medical AI³⁹Research topics in empirical studies on medical AI ethics⁴⁰

Keyword cluster analysis

Keyword co-occurrence networks can be further analysed through clustering to identify distinct research themes. Based on network structure and clustering validity, CiteSpace generates Modularity Q and Mean Silhouette S values, which are key metrics for evaluating cluster quality. Generally, Q > 0.3 indicates a significant community structure. S = 0.7 signifies high clustering reliability, and values S > 0.5 are generally considered reasonable.⁴¹ This study yielded nine Chinese clusters (Q = 0.5502, S = 0.8943) and eight English clusters (Q = 0.4583, S = 0.7222), both of which demonstrate significant network community structures and high clustering reliability. Detailed information is presented in Table 6. The clustering results reveal that global research attention focuses on several core themes: “Clinical Applications of AI Technologies”, “AI in Medical Education”, “AI in Biological Heredity”, “Ethical Principles and Issues”, and “Turing Test” globally.

Table 6.

Cluster analysis of keywords.

Cluster number	Keywords (CNKI)			Keywords (WOSCC)
Cluster number	Cluster label	Node value	Silhouette value	Cluster label	Node value	Silhouette value
#0	Medical and Health	35	0.940	AI ethics	60	0.746
#1	Medical Ethics	21	0.930	Artificial intelligence	58	0.725
#2	Ethical Principles	17	0.904	Education & educational research	57	0.819
#3	Ethical Issues	15	0.721	Social issues	51	0.7
#4	Ethics	14	0.828	Computer science	50	0.709
#5	Medical Education	7	0.896	Deep learning	49	0.765
#6	Clinical Applications	7	0.943	Medical informatics	43	0.5
#7	Turing Test	4	0.988	Genetics & heredity	8	0.99
#8	Privacy	3	0.983

Choose keyword sources to label clusters. CNKI: ① Timespan: 2017–2025 (Slice Length = 1); ② Selection Criteria: g-index (k = 25), LRF = 3.0, L/N = 10, LBY = 5, e = 1.0; ③ Network: N = 153, E = 297 (Density = 0.0255); ④ Largest 1 CCs: 123 (80%); ⑤ Nodes Labelled: 1.0%; ⑥ Pruning: None; ⑦ Modularity Q = 0.5502; ⑧ Weighted Mean Silhouette S = 0.8943; ⑨ Harmonic Mean (Q, S) = 0.6813. WOSCC: ① Timespan: 1999–2025 (Slice Length = 1); ② Selection Criteria: g-index (k = 25), LRF = 3.0, L/N = 10, LBY = 5, e = 1.0; ③ Network: N = 392, E = 1557 (Density = 0.0203); ④ Largest 1 CCs: 376 (95%); ⑤ Nodes Labeled: 1.0%; ⑥ Pruning: None; ⑦ Modularity Q = 0.4583; ⑧ Weighted Mean Silhouette S = 0.7222; ⑨ Harmonic Mean (Q, S) = 0.5607.

Analysis of keyword burstiness

Keyword burst detection identifies emerging research hotspots and reveals thematic evolution over time. The results show a dynamic, phase-based evolution of research focuses. With a burst threshold set at 0.21, 14 high-burst keywords were identified in CNKI. The evolution of its research themes can be divided into four periods: 1) Hot keywords included “medical devices”, “medical safety”, “ethical challenge”, and “doctor-patient relationship”, indicating that ethical challenges caused by medical devices, particularly those straining doctor-patient relationships, became a core research focus starting in 2019. 2) By 2020, “big data”, “ethical design”, and “ethics” gained prominence, reflecting how the rise of big data drew scholarly attention to ethical design. 3) Since 2021, with the emergence of intelligent medicine and smart medicine, research has focused primarily on technology ethics and ethical principles. 4) From 2022 to the present, ethical dilemmas and practical applications of AI in healthcare have continued to attract significant attention (Figure 4(a)).

Figure 4.

Keyword burst analysis (a) keyword burst analysis in CNKI, threshold at 0.21; (b) keyword burst analysis in WOSCC, threshold at 0.37.

For WOSCC studies, a burst threshold of 0.37 identified 14 high-burst keywords. The thematic evolution unfolded in three phases: 1) From 2017 to 2020, “history & philosophy of medicine”, “machine learning”, “precision medicine”, “digital medicine”, “data science”, “big data”, and “care robots” served as hot keywords. As the theoretical foundation of AI was deeply explored, scholars increasingly focused on data-driven medicine. The emergence of precision medicine, digital medicine, and care robots signalled trends toward personalisation, remote care, and automation in medicine. 2) From 2021 onward, attention shifted to the clinical application of AI-driven devices and the associated ethical issues. 3) By 2022, “digital twins” and “decision support systems” signalled a deepening of research into AI technologies within virtual and complex intelligent systems (Figure 4(b)).

Discussion

Ethical issues have emerged as global hotspots in the medical application of AI. This study found that research hotspots in this area encompass education, clinical practice, and academia. Research trends are shifting from identifying and governing ethical issues to designing and optimising complex AI tools embedded with ethical principles.

Descriptive analysis identifies commonalities and disparities among studies from different databases across temporal, journal, national/institutional, and authorial dimensions. Regarding publication time, 2017 marks a pivotal year in this research field. In 2016, the news that Google's deep learning-based AI AlphaGo's victory over South Korean 9th Dan Go champion Lee Sedol caused a sensation in both fields of AI and Go, which marked a significant milestone in the development of AI.⁴² The following year, top international journals successively published articles related to medical AI.^43,44 Since then, ethical issues pertaining to medical AI have received widespread and sustained scholarly attention.⁴⁵

Analysis of national and institutional contributions reveals distinct publication patterns. The USA and Canada are the most productive countries in terms of overall output in the WOSCC database. A significant volume of literature is also published in Chinese and indexed in the CNKI database. Additionally, the research landscape in this field is characterised by a wide distribution of contributing teams and institutions, which is common in rapidly evolving interdisciplinary areas. While this demonstrates broad interest, it also suggests that the field may still be in a phase of growing consolidation. Encouraging sustained collaboration both within and across national research communities could be instrumental in fostering deeper integration and advancing the field collectively.

To map the research landscape, we further reviewed and inductively categorised the highly cited papers and key articles associated with these keywords into four main perspectives: 1) ethical issues in different application scenarios of AI technique in clinical practice, 2) AI ethical concerns and considerations in medical academia, 3) ethical challenges of AI in medical education, 4) a macroscopic perspective discussing ethical governance, regulation and review. To further understand these research hotspots, an ethical framework could be helpful.¹⁰

The technology is applied across clinical specialities, including ophthalmology,⁴⁶ imaging,⁴⁷ neurology,⁴⁸ gastroenterology,⁴⁹ paediatrics,⁵⁰ and mental health,⁵¹ supporting tasks from image analysis to diagnostic consultation. When deploying the technology across these fields, we must carefully consider the associated ethical challenges, encompassing issues of algorithmic transparency,⁵² justice and fairness, patient consent,⁵³ privacy, and responsibility.⁵⁴ These can be categorised into three ethical dimensions: epistemic, normative, and traceability ethics.¹⁰ Epistemic concerns centre on the need for explainability and accuracy. Although explainability promotes trust and perceived fairness,⁵⁵ opaque decision-making remains commonplace in clinical practice, and diagnostic accuracy may take precedence over interpretability in certain contexts.⁵⁶ Normative ethics underscore the necessity of strengthened regulation, particularly regarding fairness, as biases can originate from data, algorithm design, and clinical interactions.⁵⁷ Beyond improvements in data management and clinician-patient education, addressing these requires algorithmic auditing and a redefinition of fairness in technology and clinical settings.⁵⁸ Simultaneously, privacy protection requires dynamic strategies, such as technologically enabled recurrent informed consent, clear data withdrawal rights, and evolving data protection, anonymisation, and regulation.⁵⁹ Within traceability ethics, a key challenge resides in determining ultimate accountability,⁶⁰ which in healthcare often remains permissible only when the technology serves a strictly auxiliary and supplementary role.⁵⁰ These ethical domains are deeply interrelated and cannot be resolved in isolation. Ultimately, navigating them may require a careful balance to achieve clinical and ethical acceptance.

The use of AI in medical academia has likewise raised significant ethical concerns and considerations.^61,62 These relate largely to research misconduct, including issues within funded projects,⁶³ academic writing and publication practices.⁶⁴ Key problems involve information inaccuracy and fabrication, touching on epistemic and traceability ethics.^10,65 Furthermore, while AI has the potential to bridge language barriers in global research, it also risks widening inequalities if primarily leveraged by high-income countries and privileged researchers (normative ethics).⁶⁶ In response, scholars have conducted extensive studies covering attitude surveys, strategic proposals, publishing guidelines,⁶¹ critical discussions,⁶⁵ and ethical reflections.⁶⁷ Addressing these issues will require strengthening education and guidance for researchers, advancing detection technologies for AI-generated content, and coordinating international regulatory efforts.⁶⁸

Turning to medical education, ethical challenges posed by AI include privacy and data security, overreliance on technology, a tendency toward utilitarian teaching, and potential impacts on doctor–patient relationships.^51,69 These issues largely fall within epistemic and normative ethical domains.¹⁰ Current research exhibits notable regional divergence. Studies indexed in CNKI predominantly review the current state of AI adoption in medical schools, analysing associated ethical challenges and responses, with a focus on leveraging AI to foster professional competencies and ethical literacy.⁶⁹ In contrast, literature in WOSCC demonstrates broader thematic diversity, including the development of scales to assess readiness for medical AI,⁷⁰ debates on the necessity of AI ethics education,⁷¹ the formulation of application guidelines,⁷² and the exploration of innovative teaching methods.⁷³ This divergence highlights the need for developing countries to assess the global significance of AI ethics education, integrate it with the realities of their medical education landscape, and develop teaching approaches and systems tailored to national needs. Furthermore, greater attention should be directed toward understanding the needs and acceptance of AI ethics education among medical faculty and students, as such insights are essential for guiding curricular reform and implementing effective strategies.

From a macro governance perspective, the growing complexity of ethical challenges underscores the need for robust oversight frameworks to ensure the safe, effective, and equitable application of AI. Research indexed in CNKI primarily adopts a review-based approach, identifying core governance elements: privacy protection, algorithmic transparency and justice, clear accountability, and the enhancement of public trust.⁷⁴ Corresponding recommendations emphasise strengthening legal and regulatory systems, expanding educational initiatives, and cultivating professional expertise.^74–77 Similar themes are also explored in WOSCC literature.^78,79 Some studies further underscore persistent implementation gaps. Although practical guidelines exist for specialised fields,^80,81 a significant chasm remains between ethical principles and their operationalisation, largely attributable to the lack of a clear, actionable definition of these principles.⁸² Establishing clear operational definitions is crucial to mitigate the risks of unintentional misunderstandings of the term, as well as intentional misuse/ethics washing. However, we may caution that over-technicalising the definitions of ethical principles may conflict with the interpretive, relational, and normative dimensions of medicine, which cannot be codified.⁸³

Strengths and limitations

This study employs bibliometric analysis to identify research trends and emerging topics in AI ethics within the medical field. Given the broad coverage of research domains and wide adoption in previous bibliometric studies, we selected the WOSCC database. To mitigate language bias and capture research from China, we also included the CNKI database, a comprehensive and authoritative resource in the region. Analysing these databases separately allowed for comparative insights into content similarities and differences. Furthermore, visualisations of research trends were generated to facilitate a clear and efficient understanding of the landscape and to highlight potential future research directions.

However, notwithstanding its systematic approach, this study has several limitations that should be acknowledged. Primarily, our reliance on only two databases may give rise to selection bias, as this omits relevant studies indexed in other databases (e.g., PubMed). Secondly, this study adopts a macro perspective to map the research landscape, and a more granular analysis of research hotspots within specific subareas was precluded. Thirdly, we delimit publication types to research papers, excluding editorials, commentaries, viewpoints, etc., which may reduce the comprehensiveness of the results. Additionally, as a bibliometric study, this work does not assess the quality of the included literature. Consequently, our analysis depends on assessments of content relevance, a process inherently influenced by the researcher's expertise and judgement, which may limit the objectivity and replicability of the results. Future research could address these limitations by expanding database coverage, incorporating diverse publication types, and integrating more objective indicators or automated tools to enhance analytical rigour and reproducibility.

Conclusion

Academic attention to AI ethics in the medical field has risen markedly since 2017, encompassing diverse research perspectives and methodological approaches, gradually forming a systematic, multidimensional research system. Although stable journals and influential, productive countries have been established, team/country collaboration remains to be further strengthened. The shift in research trends reflects an evolution from fragmentation to integration of technology and ethics in medicine. Future research should further refine the operational definitions of ethical principles across diverse medical AI application scenarios, thereby guiding researchers to design medical AI systems with in-depth integration of technology and ethics, and also providing a clear foundation for educators to design and implement relevant courses and training, as well as for policymakers to formulate enforceable ethical norms and regulatory standards.

Supplemental Material

sj-docx-1-dhj-10.1177_20552076261431438 - Supplemental material for Research hotspots and trends of artificial intelligence ethics in medicine: A bibliometric analysis from 1999 to 2025

Supplemental material, sj-docx-1-dhj-10.1177_20552076261431438 for Research hotspots and trends of artificial intelligence ethics in medicine: A bibliometric analysis from 1999 to 2025 by Jialu Li, Jun Li, Yunjing Qiu, Shiquan He, Zhenxiang Zhang, Peng Wang, Hui Zhang and Beilei Lin in DIGITAL HEALTH

Supplemental Material

sj-docx-2-dhj-10.1177_20552076261431438 - Supplemental material for Research hotspots and trends of artificial intelligence ethics in medicine: A bibliometric analysis from 1999 to 2025

Supplemental material, sj-docx-2-dhj-10.1177_20552076261431438 for Research hotspots and trends of artificial intelligence ethics in medicine: A bibliometric analysis from 1999 to 2025 by Jialu Li, Jun Li, Yunjing Qiu, Shiquan He, Zhenxiang Zhang, Peng Wang, Hui Zhang and Beilei Lin in DIGITAL HEALTH

Footnotes

List of abbreviations

Acknowledgements

We would like to thank all team members for their indispensable assistance in conducting this research.

ORCID iDs

Jialu Li

Beilei Lin

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Authors’ contributions

Jialu LI: Conceptualization, Methodology, Formal analysis, Writing - Original Draft, Visualization; Jun LI: Writing - Review & Editing; Yunjing Qiu: Writing - Review & Editing; Shiquan HE: Visualization; Zhenxiang Zhang: Supervision, Writing - Review & Editing; Peng Wang: Supervision, Funding acquisition; Hui Zhang: Writing - Review & Editing; Beilei LIN: Conceptualization, Writing - Review & Editing, Supervision, Funding acquisition. All authors have approved the submitted version and are accountable for the contributions and integrity of the paper.

Funding

This work was supported by the scholarships from the Postdoctoral Research Foundation of China [2023M743197], Foundation of Henan Educational Committee [25A320027], and National Natural Science Foundation of China [72104221, 72274180].

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Availability of data and materials

The data that support the findings of this study are available on reasonable request from the corresponding author.

Patient and public involvement

Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

Trial registration

Not applicable.

Supplemental material

Supplemental material for this article is available online.

References

Kaul

Enslin

Gross

. History of artificial intelligence in medicine. Gastrointest Endosc 2020; 92: 807–812.

Montejo

Fenton

Davis

. Artificial intelligence (AI) applications in healthcare and considerations for nursing education. Nurse Educ Pract 2024; 80: 104158.

Yin

Ngiam

Teo

. Role of artificial intelligence applications in real-life clinical practice: systematic review. J Med Internet Res 2021; 23: e25759.

Alowais

Alghamdi

Alsuhebany

, et al. Revolutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Med Educ 2023; 23: 689.

Yue

. Multidimensional dilemma of medical affective computing and its legal realization. Med Philos 2022; 43: 39–44.

Alhussain

Kelly

O'Flaherty

, et al. Emerging role of artificial intelligence in global health care. Health Policy Technol 2022; 11: 100661.

Wei

, et al. Innovation and challenges of artificial intelligence technology in personalized healthcare. Sci Rep 2024; 14: 18994.

Chen

Cheng

. Algorithm black box in medical artificial intelligence and its core ethical issues. Med Philos 2024; 45: 6–10.

Ouyang

Yang

, et al. Research on privacy protection in the context of healthcare data based on knowledge map. Medicine (Baltimore) 2024; 103: e39370.

10.

Morley

Machado

CCV

Burr

, et al. The ethics of AI in health care: a mapping review. Soc Sci Med 2020; 260: 113172.

11.

Tran

, et al. Global evolution of research in artificial intelligence in health and medicine: a bibliometric study. J Clin Med 2019; 8: 360.

12.

Shi

Bendig

Vollmar

, et al. Mapping the bibliometrics landscape of AI in medicine: methodological study. J Med Internet Res 2023; 25: e45815.

13.

Cui

Chang

Yang

, et al. Development and trends in artificial intelligence in critical care medicine: a bibliometric analysis of related research over the period of 2010-2021. J Pers Med 2023; 13: 50.

14.

Sun

Liu

, et al. Mapping key nodes and global trends in AI and large language models for medical education: a bibliometric study. Adv Med Educ Pract 2025; 16: 1421–1438.

15.

Roy

Shah

Ahluwalia

, et al. Analyzing the evolution of medical ethics education: a bibliometric analysis of the top 100 cited articles. Cureus 2023; 15: e41411.

16.

Wang

Sun

. Development trends of ethical research on artificial intelligence in medicine: topic distribution, knowledge base and future prospects. J Dial Nat 2023; 45: 18–29.

17.

Saheb

Carpenter

. Mapping research strands of ethics of artificial intelligence in healthcare: a bibliometric and content analysis. Comput Biol Med 2021; 135: 104660.

18.

Gulkesen

Sonuvar

. Data privacy in medical informatics and electronic health records: a bibliometric analysis. Health Care Anal 2025. DOI: 10.1007/s10728-025-00519-0.

19.

Yan

Sun

. Research on the application of artificial intelligence in the field of biomedical treatment and the coping strategies of legal and ethical issues. Sci-tech Finance Monthly 2017: 55–60. DOI: https://doi.org/10.26916/j.cnki.stfmonthly.2017.03.025.

20.

Kluge

. Medical narratives and patient analogs: the ethical implications of electronic patient records. Methods Inf Med 1999; 38: 253–259.

21.

Chen

. CiteSpace: text mining and visualization in scientific literature. Beijing: Capital University of Economics and Business Press, 2022.

22.

Chen

. Citespace II: detecting and visualizing emerging trends and transient patterns in scientific literature. J Assoc Inf Sci Technol 2006; 57: 359–377.

23.

Kasun

Ryan

Paik

, et al. Academic machine learning researchers’ ethical perspectives on algorithm development for health care: a qualitative study. J Am Med Inform Assoc 2024; 31: 563–573.

24.

Hesjedal

Lyso

Solbjor

, et al. Valuing good health care: how medical doctors, scientists and patients relate ethical challenges with artificial intelligence decision-making support tools in prostate cancer diagnostics to good health care. Sociol Health Illn 2024; 46: 1808–1827.

25.

Čartolovni

Malešević

Poslon

. Critical analysis of the AI impact on the patient-physician relationship: a multi-stakeholder qualitative study. Digit Health 2023; 9: 20552076231220833.

26.

Gundlack

Negash

Thiel

, et al. Artificial intelligence in medical care - patients’ perceptions on caregiving relationships and ethics: a qualitative study. Health Expect 2025; 28: e70216.

27.

Goirand

Austin

Clay-Williams

. Mapping the flow of knowledge as guidance for ethics implementation in medical AI: a qualitative study. PLoS One 2023; 18: e0288448.

28.

Petersson

Vincent

Svedberg

, et al. Ethical considerations in implementing AI for mortality prediction in the emergency department: linking theory and practice. Digit Health 2023; 9: 20552076231206588.

29.

Koçer Tulgar

Tulgar

Güven Köse

, et al. Anesthesiologists’ perspective on the use of artificial intelligence in ultrasound-guided regional anaesthesia in terms of medical ethics and medical education: a survey study. Eurasian J Med 2023; 55: 146–151.

30.

Al-Ani

Rayyan

Maswadeh

, et al. Evaluating the understanding of the ethical and moral challenges of Big Data and AI among Jordanian medical students, physicians in training, and senior practitioners: a cross-sectional study. BMC Med Ethics 2024; 25: 18.

31.

Wang

, et al. Comparative survey among paediatricians, nurses and health information technicians on ethics implementation knowledge of and attitude towards social experiments based on medical artificial intelligence at children's hospitals in Shanghai: a cross-sectional study. BMJ Open 2023; 13: e071288.

32.

Mlakar

Kampič

Flis

, et al. Study protocol: a survey exploring patients’ and healthcare professionals’ expectations, attitudes and ethical acceptability regarding the integration of socially assistive humanoid robots in nursing. BMJ Open 2022; 12: e054310.

33.

Zhang

Yin

, et al. Development, reliability, and structural validity of the scale for knowledge, attitude, and practice in ethics implementation among AI researchers: cross-sectional study. JMIR Form Res 2023; 7: e42202.

34.

Martinho

Kroesen

Chorus

. A healthy debate: exploring the views of medical doctors on the ethics of artificial intelligence. Artif Intell Med 2021; 121: 102190.

35.

Balas

Wadden

Hébert

, et al. Exploring the potential utility of AI large language models for medical ethics: an expert panel evaluation of GPT-4. J Med Ethics 2024; 50: 90–96.

36.

Sridharan

Sivaramakrishnan

. Leveraging artificial intelligence to detect ethical concerns in medical research: a case study. J Med Ethics 2025; 51: 126–134.

37.

Faqar-Uz-Zaman

Filmann

Mahkovic

, et al. Study protocol for a prospective, double-blinded, observational study investigating the diagnostic accuracy of an app-based diagnostic health care application in an emergency room setting: the eRadaR trial. BMJ Open 2021; 11: e041396.

38.

Willems

Bak

Tan

, et al. Ethical issues in two parallel trials of personalised criteria for implantation of implantable cardioverter defibrillators for primary prevention: the PROFID project-a position paper. Open Heart 2021; 8: e001686.

39.

Grote

. Randomised controlled trials in medical AI: ethical considerations. J Med Ethics 2022; 48: 899–906.

40.

Tang

Fantus

. Medical artificial intelligence ethics: a systematic review of empirical studies. Digit Health 2023; 9: 20552076231186064.

41.

Chen

C-M

Liu

Z-Y

, et al. Methodological function of CiteSpace’s knowledge graph. Studies in Science of Science 2015; 33: 242–253.

42.

Silver

Huang

Maddison

, et al. Mastering the game of Go with deep neural networks and tree search. Nature 2016; 529: 484–489.

43.

Esteva

Kuprel

Novoa

, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature 2017; 542: 115–118.

44.

Gulshan

Peng

Coram

, et al. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA 2016; 316: 2402–2410.

45.

Yuste

Goering

Arcas

BAY

, et al. Four ethical priorities for neurotechnologies and AI. Nature 2017; 551: 159–163.

46.

Abdullah

Schuman

Shabsigh

, et al. Ethics of artificial intelligence in medicine and ophthalmology. Asia Pac J Ophthalmol (Phila) 2021; 10: 289–298.

47.

Abràmoff

Cunningham

Patel

, et al. Foundational considerations for artificial intelligence using ophthalmic images. Ophthalmology 2022; 129: e14–e32.

48.

Bösel

Mathur

Cheng

, et al. AI and neurology. Neurol Res Pract 2025; 7: 11.

49.

Aggarwal

Drew

Parikh

, et al.

Ethical implications of artificial intelligence in gastroenterology: the co-pilot or the captain?

Dig Dis Sci 2024; 69: 2727–2733.

50.

Allen

Earp

Wilkinson

. AI-assisted consent in paediatric medicine: ethical implications of using large language models to support decision-making. J Med Ethics 2025. DOI: https://doi.org/10.1136/jme-2024-110624.

51.

Blease

Kharko

Annoni

, et al. Machine learning in clinical psychology and psychotherapy education: a mixed methods pilot survey of postgraduate students at a Swiss university. Front Public Health 2021; 9: 623088.

52.

Durán

Jongsma

. Who is afraid of black box algorithms? On the epistemological and ethical basis of trust in medical AI. J Med Ethics 2021; 47: 329–335.

53.

Weiner

Dankwa-Mullan

Nelson

, et al. Ethical challenges and evolving strategies in the integration of artificial intelligence into clinical practice. PLoS Digit Health 2025; 4: e0000810.

54.

Corfmat

Martineau

Régis

. High-reward, high-risk technologies? An ethical and legal account of AI development in healthcare. BMC Med Ethics 2025; 26: 4.

55.

Angerschmid

Zhou

Theuermann

, et al. Fairness and explanation in AI-informed decision making. Mach Learn Know Extr 2022; 4: 556–579.

56.

London

. Artificial intelligence and black-box medical decisions: accuracy versus explainability. Hastings Cent Rep 2019; 49: 15–21.

57.

Ueda

Kakinuma

Fujita

, et al. Fairness of artificial intelligence in healthcare: review and recommendations. Jpn J Radiol 2024; 42: 3–15.

58.

Liu

Ning

Teixayavong

, et al. A translational perspective towards clinical AI fairness. NPJ Digit Med 2023; 6: 172.

59.

Murdoch

. Privacy and artificial intelligence: challenges for protecting health information in a new era. BMC Med Ethics 2021; 22: 122.

60.

Allen

Earp

Koplin

, et al.

Consent-GPT: is it ethical to delegate procedural consent to conversational AI?

J Med Ethics 2024; 50: 77–83.

61.

Ahn

. Large language model usage guidelines in Korean medical journals: a survey using human-artificial intelligence collaboration. J Yeungnam Med Sci 2025; 42: 14.

62.

Doyal

Sender

Nanda

, et al. ChatGPT and artificial intelligence in medical writing: concerns and ethical considerations. Cureus 2023; 15: e43292.

63.

Zhang

. Research on privacy issues in artificial intelligence research projects. China Modern Doctor 2022; 60: 109–111.

64.

Jiang

Sun

. The application of artificial intelligence in the editing and publishing of medical journals. Sci-Technol Publ 2020: 66–71. DOI: https://doi.org/10.16510/j.cnki.kjycb.20200218.008.

65.

Currie

. Academic integrity and artificial intelligence: is ChatGPT hype, hero or heresy? Semin Nucl Med 2023; 53: 719–730.

66.

Dergaa

Chamari

Zmijewski

, et al. From human writing to artificial intelligence generated text: examining the prospects and potential threats of ChatGPT in academic writing. Biol Sport 2023; 40: 615–622.

67.

Lei

Dong

, et al. Thoughts on medical journals dealing with medical ethics under the background of artificial intelligence. Acta Editologica 2023; 35: 263–267.

68.

Cheng

Wang

Lei

, et al. Emerging challenges in research integrity governance driven by Al: an analysis of global retractions. Chin J Med Sci Res Manage 2025; 38: 462–468.

69.

Yao

Hou

Zhang

, et al. The application and exploration of artificial intelligence in the cultivation of professional skills and ethics qualities of medical talents. Chin Med Ethics 2019; 32: 1477–1480.

70.

Karaca

Çaliskan

Demir

. Medical artificial intelligence readiness scale for medical students (MAIRS-MS) - development, validity and reliability study. BMC Med Educ 2021; 21: 112.

71.

Katznelson

Gerke

. The need for health AI ethics in medical school education. Adv Health Sci Educ Theory Pract 2021; 26: 1447–1458.

72.

Masters

. Ethical use of artificial intelligence in health professions education: AMEE guide No.158. Med Teach 2023; 45: 574–584.

73.

Weidener

Fischer

. Proposing a principle-based approach for teaching AI ethics in medical education. JMIR Med Educ 2024; 10: e55368.

74.

Shen

Jiang

, et al. Research on the focus of ethical governance in the field of artificial intelligence in medicine. Chin Med Ethics 2024; 37: 39–44.

75.

Song

Huo

Deng

. Problems, challenges and countermeasures of ethical governance of science and technology in China. Stud Sci Sci 2024; 42: 1569–1576+1595.

76.

Zhang

. Discrimination of the governance principles of science and technology ethics and dissemination of ideas. Chin Med Ethics 2022; 35: 479–482.

77.

Jiang

Chen

. Construction of ethical review system of artificial intelligence application under the context of health and medical big data. Chin Med Ethics 2020; 33: 841–846.

78.

McKay

Williams

Prestwich

, et al. Artificial intelligence and medical research databases: ethical review by data access committees. BMC Med Ethics 2023; 24: 49.

79.

Herington

McCradden

Creel

, et al. Ethical considerations for artificial intelligence in medical imaging: deployment and governance. J Nucl Med 2023; 64: 1509–1515.

80.

Zsidai

Hilkert

Kaarre

, et al. A practical guide to the implementation of AI in orthopaedic research - part 1: opportunities in clinical application and overcoming existing challenges. J Exp Orthop 2023; 10: 117.

81.

Iserson

. Informed consent for artificial intelligence in emergency medicine: a practical guide. Am J Emerg Med 2024; 76: 225–230.

82.

Bürger

Amann

Bui

CKT

, et al. The unmet promise of trustworthy AI in healthcare: why we fail at clinical translation. Front Digit Health 2024; 6: 1279629.

83.

Sirago

Solarino

Dell'Erba

, et al. Medicine as a human science: not everything that counts can be codified. Health Care Anal 2026. DOI: https://doi.org/10.1007/s10728-025-00556-9.