The experience of nursing students using generative artificial intelligence: a qualitative meta-synthesis

Abstract

Objective

This study systematically integrated qualitative research regarding nursing students’ experiences using Generative Artificial Intelligence (GAI). It aimed to provide an evidence-based basis to facilitate the safe, appropriate, and effective use of GAI among nursing students.

Design

A qualitative systematic review.

Data sources

PubMed, Web of Science, The Cochrane Library, CINAHL, Embase, CNKI, Wanfang, VIP, and SinoMed were searched for qualitative studies addressing nursing students’ experiences with GAI, covering the period from January 1, 2015, to January 29, 2026.

Methods

The methodological quality of included studies was evaluated using qualitative research quality appraisal criteria developed by the JBI Centre for Evidence-Based Health Care (Australia). A meta-synthesis approach was used to synthesize findings, and the ConQual tool was applied to assess confidence in the results.

Results

Twenty-four studies were included, producing 41 findings grouped into 11 sub-themes and subsequently synthesized into four themes: perceived benefits, perceived shortcomings, perceived risks, and perceived needs.

Conclusions

Nursing students’ experiences with GAI were found to be complex and sometimes contradictory. Therefore, educational institutions, educators, and technology developers should enhance relevant technical, educational, and policy support frameworks to ensure the appropriate, safe, and effective use of GAI tools among nursing students.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12912-026-04663-3.

Introduction

Generative Artificial Intelligence (GAI) integrates machine learning, natural language processing, reasoning, and decision-making within a unified framework. It generates original content that aligns with user-defined styles based on complex prompts and data processed through deep learning techniques [1], including text, images, audio, and video [2]. The release of ChatGPT in 2022 was widely regarded as a milestone that initiated a new wave of artificial intelligence innovation. In recent years, driven by continuous technological progress, a series of large language models, including DeepSeek and Gemini, have emerged [3]. Owing to its advanced language processing capabilities, GAI has been widely adopted across various sectors, with particularly strong applications in healthcare, education, and training [3, 4].

Nursing, a discipline that requires substantial theoretical knowledge and high-level practical competence, has experienced significant changes in educational approaches following the emergence of GAI. For educators, GAI has been increasingly used as a supportive resource for curriculum design and instructional material development [5]. For learners, this technology serves as an adjunct to instruction and practice by enhancing learning efficiency, expanding knowledge acquisition pathways, and supporting academic writing [6]. In addition, personalized and interactive learning experiences can be provided through realistic patient scenarios for skills training [7]. From a broader educational perspective, GAI can facilitate the modernization of nursing workforce training models by simulating clinical decision-making processes [8]. However, rapid technological advancement has also introduced new challenges. Previous research reported that ChatGPT achieved an overall accuracy of 80.75% in the nursing examinations in Taiwan, raising concerns regarding academic integrity [9]. Furthermore, the use of GAI may foster technological dependence among students, potentially weakening their ability to evaluate information quality and apply critical thinking [10]. More importantly, because GAI systems lack direct access to the latest medical research and real-time clinical data, accurate and up-to-date information may not always be ensured, and the possibility of disseminating inaccurate or misleading content cannot be ruled out [5]. The rapid evolution of artificial intelligence represents an irreversible technological transformation [11], with GAI playing a central role in educational modernization and personalized learning [12]. Therefore, optimizing the effectiveness of this tool is imperative.

As a key component of nursing education and an important user group of intelligent technologies, nursing students’ experiences and perceptions of GAI use must be systematically examined to maximize its benefits. Qualitative research is considered a suitable method for exploring nursing students’ experiences with GAI tools. A systematic literature search indicated that original studies investigating nursing students’ subjective experiences with GAI have gradually increased. However, no prior qualitative meta-synthesis has focused specifically on nursing students’ experiences with GAI tools. In addition, heterogeneity related to small sample sizes, methodological differences, and diverse research contexts limits the generalizability of individual studies. Moreover, many qualitative studies focus on specific populations or particular GAI tools, which restricts their ability to represent the broader and shared experiences of nursing students across multiple GAI platforms.

Therefore, the present study is a meta-synthesis of qualitative research examining nursing students’ experiences with GAI tools. This review was conducted to provide evidence-based support for the development of GAI usage guidelines and ethical standards in nursing education institutions. It also aims to offer a theoretical basis for educators seeking to establish scientifically grounded approaches for integrating GAI into nursing curricula. Furthermore, potential directions for optimizing GAI functionality are proposed for technology developers. Through these efforts, nursing students’ artificial intelligence literacy may be enhanced, thereby promoting more effective engagement with rapid technological advancements.

The research question was: What are nursing students’ experiences with the use of GAI (e.g., ChatGPT) ?

Methods

Design

This study performed a meta-synthesis of qualitative studies exploring nursing students’ experiences and perceived needs related to the use of GAI tools. This study followed the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [13] and the Enhancing Transparency in Reporting the Synthesis of Qualitative Research (ENTREQ) [14]. A systematic evaluation of qualitative evidence was conducted using the JBI methodology [15], with the JBI Critical Appraisal Checklist for Qualitative Research used for quality assessment. The review was registered with PROSPERO (CRD420251011437).

Research team and reflexivity

The research team consisted of two nursing education experts, both experienced in qualitative research. The first author specialized in artificial intelligence research in nursing during their degree program and possessed considerable expertise in this field. Regular meetings were held to discuss and reflect upon potential biases and evaluate their possible impact on the research. This reflective process aimed to ensure a more objective and impartial interpretation of the study findings.

Search strategy

The search included PubMed, Web of Science, The Cochrane Library, CINAHL, Embase, CNKI, Wanfang, VIP, and SinoMed, covering qualitative studies on nursing students’ experiences with GAI from January 1, 2015, to January 29, 2026. A combination of subject terms and free-text keywords was used, and relevant references were manually traced. The complete search strategy is provided in Supplementary Material 1. Search terms: “nurs* student*”, “pupil nurs*”, “Generative Artificial Intelligence”, “Generative AI”, “GenAI”, “Chatbot*”, “ChatGPT”, “Chat GPT”, “Chat-GPT”, “Large Language Model*”. Throughout the retrieval process, composition was done using boolean operators. To ensure comprehensiveness, the search was not restricted by study type.

Inclusion and exclusion criteria

Inclusion and exclusion criteria were developed based on the PICoS framework recommended by the JBI Centre for Evidence-Based Healthcare.

Inclusion criteria: (1) Population (P): nursing students; (2) Phenomenon of Interest (I): experiences and perceptions of using GAI; (3) Context (Co): use of GAI in any setting; (4) Study Design (S): qualitative studies or qualitative components of mixed-methods studies.

Exclusion criteria: (1) Non-English or non-Chinese publications; (2) Studies with unavailable full text or incomplete information; (3) Studies that were rated as quality grade C according to the Joanna Briggs Institute (JBI) quality assessment criteria.

Screening and data extraction

All retrieved literature was imported into EndNote 21, and duplicates were removed. Two researchers independently screened the titles, abstracts, and full texts according to the predefined inclusion and exclusion criteria. The results were cross-checked. The specific steps of this process are shown in Fig. 1. Two researchers independently extracted data, including the first author, publication year, country, research method, sample size, context, phenomenon of interest, and main findings. For studies employing a mixed-methods design, based on the aforementioned criteria, only data from the qualitative component were extracted, and data from the quantitative component were not extracted. In this context, the sample size refers to the number of participants involved in qualitative data collection. Any discrepancies were resolved through discussion with a third researcher. Table 1 shows the final results.

Fig. 1.

Flowchart of literature screening

Table 1.

Basic characteristics of the included studies (n = 24)

Author (year)	Country	sample	Methodologies	Aim	Findings
Ma et al. (2024) [18]	China	15	Descriptive qualitative research; semi-structured interviews	To understand nursing students’ experiences using large language model-based chatbots.	Four themes emerged: 1. Diverse experiences of use 2. Differentiated usage scenarios 3. Potential issues related to use 4. Self-perceived needs and expectations
Liu et al. (2024) [19]	China	14	Descriptive qualitative research; semi-structured interviews	To understand Master’s-level nursing students’ experiences using GAI tools.	Three themes emerged: 1. High interest in GAI tools 2. Perceived benefits from using GAI tools 3. Concerns regarding the use of GAI tools
Ma et al. (2025) [20]	China	24	Descriptive qualitative research; semi-structured interviews	To understand nursing students’ experiences regarding the use of large language models, including facilitating factors and barriers.	Three themes emerged: 1. Attitude 2. Subjective norm 3. Perceived behavioral control
Hawk et al. (2024) [21]	US	19	Descriptive qualitative research; reflective thematic analysis	To explore nursing students’ views on the use of GAI chatbots for answering clinical questions.	Four themes emerged: 1. Surprisingly familiar 2. The importance of critical thinking and external validation 3. A good summary that lacks depth and nuance 4. Cautiously optimistic
Summers et al. (2024) [22]	Australia	13	Descriptive qualitative research; semi-structured interviews	To explore nursing students’ views on the application and integration of GAI tools in the learning process.	Six themes emerged: 1. Educational impact of AI tools 2. Equitable learning environment 3. Ethical considerations 4. Technology integration 5. Safety and practical utility 6. Generational differences
Gunawan et al. (2024) [23]	Thailand	19	Descriptive qualitative research; semi-structured interviews	To explore nursing students’ views on the use of ChatGPT in the learning process.	Three themes emerged: 1. ChatGPT may be helpful but not fully trusted 2. Nurse skills will never be replaced 3. The need for evaluating and advancing curriculum
Rodriguez-Arrastia et al. (2022) [24]	Spain	114	Descriptive qualitative research; semi-structured interviews	To explore final-year nursing students’ experiences and perceptions regarding the acceptability and feasibility of using chatbots for clinical decision-making and patient safety.	Two themes emerged: 1. Experiences in the use of a chatbot service for clinical decision-making 2. Integrating conversational agents into organizational safety culture
Chang et al. (2022) [25]	Taiwan, China	5	Hybrid study	To understand nursing students’ learning responses to mobile chatbot-based learning methods.	Four themes emerged: 1. User-friendly interface 2. Enhancing learning engagement 3. Enhancing self-efficacy 4. Other feedback
Reading Turchioe et al. (2024) [26]	US	10	Hybrid study	To explore nursing students’ perceptions of the security and usability of ChatGPT.	Two themes emerged: 1. Reflections on safety 2. Reflections on usability
Zheng et al. (2025) [27]	China	15	Descriptive qualitative research; semi-structured interviews	To understand Master’s-level nursing students’ views on the application of large language models.	Four themes emerged: 1. Application scenarios of large language models 2. External influencing factors 3. Challenges in large language model applications 4. Concerns regarding large language model applications
Han et al. (2025) [28]	US	99	Descriptive qualitative research; semi-structured interviews	To understand nursing students’ perceptions and views regarding GAI tools.	Four themes emerged: 1. Reasons for using GAI and assistive AI tools 2. Perceived advantages 3. Perceived disadvantages 4. Perceived support needs
Stratton-Maher et al. (2026) [29]	Australia	48	Descriptive qualitative research; focus group semi-structured interviews	To explore nursing students’ experiences, challenges, and perspectives regarding the role of ChatGPT in academic development.	Six themes emerged: 1. Navigating confusion and adaptation 2. Developing functional and academic skills through AI 3. Evaluating effectiveness and navigating reliability challenges 4. Negotiating academic integrity and ethical use of AI 5. Recognizing the benefits and future potential of AI collaboration 6. The critical role of feedback, instructional support, and digital literacy
Nelson et al. (2025) [30]	US	23	Descriptive qualitative research; focus group semi-structured interviews	To understand nursing students’ experiences and views on using AI for clinical case handling.	Four themes emerged: 1. Emotional and therapeutic appropriateness 2. Clinical accuracy and safety 3. Usefulness and practical application 4. Trust and transparency
Sağlam et al. (2025) [31]	Turkey	15	Descriptive qualitative research; semi-structured interviews	To understand nursing students’ experiences with using ChatGPT.	Four themes emerged: 1. Role in public health nursing 2. Concerns about its use 3. Benefits 4. Students’ future forecasts
Yuan et al. (2025) [32]	China	36	Hybrid study	To understand nursing students’ subjective experiences and perceptions regarding GAI tools.	Four themes emerged: 1. Strengths 2. Weaknesses 3. Opportunities 4. Threats
Aslan et al. (2025) [33]	Turkey	13	Descriptive phenomenological approach, semi-structured interviews	To explore Master’s-level nursing students’ experiences using ChatGPT to prepare seminar assignments.	Three themes emerged: 1. Opportunities 2. Hesitations 3. Recommendations for future
Shen et al. (2025) [34]	China	17	Descriptive qualitative research; semi-structured interviews	To explore nursing students’ experiences using GAI to analyze real cases.	Six themes emerged: 1. Experience with GAI usage 2. Effectiveness in nursing case analysis 3. Prompt design and its role 4. Privacy and data security 5. Emotional understanding of GAI 6. Changes in learning methods and professional perceptions
Lam et al. (2025) [35]	Canada	16	Descriptive qualitative research; focus group semi-structured interviews	To explore nursing students’ perspectives on AI tools, including perceived benefits, concerns, and challenges.	Four themes emerged: 1. Utilization as a support tool 2. Utilization leading to loss of competency in foundational skills 3. Utilization risking credibility and academic integrity 4. Need for further education and resources
Shen et al. (2025) [36]	China	16	Descriptive qualitative research; semi-structured interviews	To explore nursing students’ experiences and acceptance in using NurCaseAI for the assessment of thyroid surgery nursing cases.	Five themes emerged: 1. Perceived system capabilities and limitations 2. Learning enhancement and performance benefits 3. Interaction efficiency and usability challenges 4. Emotional engagement and human connection needs 5. Acceptance and improvement recommendations
Ahmed et al. (2025) [37]	UAE	27	Descriptive phenomenological approach, semi-structured interviews	To understand undergraduate nursing students’ diverse experiences using ChatGPT.	Three themes emerged: 1. Exploring students’ perceptions of ChatGPT 2. Students’ satisfactory and unsatisfactory experiences of ChatGPT 3. Future scope of ChatGPT use
Jembu et al. (2025) [38]	Malaysia	11	Descriptive qualitative research; semi-structured interviews	To explore nursing students’ perceptions of the use of AI tools in academic writing.	Four themes emerged: 1. Perceived strengths of AI adoption 2. Perceived flaws 3. Influential factors 4. Systemic barriers
Karaçay et al. (2025) [39]	Turkey	24	Descriptive qualitative research; semi-structured interviews	To understand nursing students’ experiences and views regarding in-class ChatGPT activities.	Three themes emerged: 1. Knowledge acquisition 2. Critical perspective 3. Satisfaction and recommendation
Siah et al. (2026) [40]	Singapore	14	Descriptive qualitative research; semi-structured interviews	To explore nursing students’ experiences with GAI in relation to academic integrity and job authenticity.	Five themes emerged: 1. Integration with learning 2. Trust and credibility of GAI information 3. Ethical considerations 4. Addressing academic integrity 5. Guidance on using GAI
Jang et al. (2026) [41]	China	16	Descriptive qualitative research; semi-structured interviews	To explore Master’s-level nursing students’ perceptions, attitudes, and demands regarding GAI.	Five themes emerged: 1. Performance expectancy 2. Effort expectancy 3. Social influence 4. Usage attitudes and behaviors 5. Boundaries to GAI

Note: Except for Ma et al. (2024), Liu et al. (2024), and Zheng et al. (2025), which were published in Chinese, all other studies were published in English

Quality appraisal

Two researchers independently assessed the methodological quality of the included studies using the Qualitative Research Quality Assessment Criteria developed by the JBI Centre for Evidence-Based Health Care [16]. A total of 10 items were evaluated, with each item marked as “yes,” “no,” or “unclear.” Studies with all 10 items rated as “yes” were considered to have an extremely low risk of bias and were assigned a rating of (A). Studies meeting some but not all quality standards were considered to have a moderate risk of bias and were assigned a rating of (B). Studies with all items rated as “no” were considered to have a high risk of bias and were assigned a rating of (C). Only studies rated A or B were included, while those rated C were excluded. For studies employing mixed-methods designs, only the qualitative component was included in the analysis. Disagreements were resolved through consultation with a third researcher. The quality assessment results are shown in Table 2.

Table 2.

Quality assessment of the included studies (n = 24)

Citations	Q1	Q2	Q3	Q4	Q5	Q6	Q7	Q8	Q9	Q10	Quality level
Ma et al. (2024) [18]	Y	Y	Y	Y	Y	N	N	Y	Y	Y	B
Liu et al. (2024) [19]	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Ma et al. (2025) [20]	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Hawk et al. (2024) [21]	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Summers et al. (2024) [22]	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Gunawan et al. (2024) [23]	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	A
Rodriguez-Arrastia et al. (2022) [24]	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Chang et al. (2022) [25]	U	Y	Y	Y	Y	N	N	Y	Y	Y	B
Reading Turchioe et al. (2024) [26]	U	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Zheng et al. (2025) [27]	Y	Y	Y	Y	Y	N	N	Y	Y	Y	B
Han et al. (2025) [28]	Y	Y	Y	Y	Y	N	N	Y	Y	Y	B
Stratton-Maher et al. (2026) [29]	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Nelson et al. (2025) [30]	Y	Y	Y	Y	Y	N	N	Y	Y	Y	B
Sağlam et al. (2025) [31]	Y	Y	Y	Y	Y	Y	N	Y	Y	Y	B
Yuan et al. (2025) [32]	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	A
Aslan et al. (2025) [33]	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	A
Shen et al. (2025) [34]	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Lam et al. (2025) [35]	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	A
Shen et al. (2025) [36]	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Ahmed et al. (2025) [37]	Y	Y	Y	Y	Y	N	N	Y	Y	Y	B
Jembu et al. (2025) [38]	Y	Y	Y	Y	Y	N	N	Y	Y	Y	B
Karaçay et al. (2025) [39]	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	A
Siah et al. (2026) [40]	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	B
Jang et al. (2026) [41]	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	A

Y, yes; N, no; U, unclear;NP, not applicable

Q1: Is there congruity between the stated philosophical perspective and the research methodology?

Q2: Is there congruity between the research methodology and the research question or objectives?

Q3: Is there congruity between the research methodology and the data collection methods?

Q4: Is there congruity between the research methodology and the representation and analysis of data?

Q5: Is there congruity between the research methodology and the interpretation of results?

Q6: Is the researcher’s cultural or theoretical positioning stated?

Q7: Is the influence of the researcher on the research, and vice versa, addressed?

Q8: Are participants and their voices adequately represented?

Q9: Is the research ethical according to current criteria, and is there evidence of ethical approval by an appropriate body?

Q10: Do the conclusions flow logically from the data analysis or interpretation?

Data synthesis

The data synthesis process had no predetermined theoretical framework. All findings were derived directly from the original findings of the included studies. The process was conducted in three stages. First, two researchers independently reviewed the selected literature, extracted relevant findings, organized the data into tables, and assigned numerical identifiers. Second, through iterative analysis, discussion, and negotiation, the researchers reformulated similar findings into standardized terminology, enabling clear conclusions. Finally, these conclusions were grouped into broader categories, which were synthesized into comprehensive findings. During data analysis, disagreements were resolved by consultation with a third researcher. The final synthesized findings were reviewed and revised, as necessary, by the entire research team.

Appraisal of synthesized findings

In this study, the ConQual approach [17] was employed to assess the quality of the synthesized findings, evaluating them across five dimensions of dependability and three dimensions of credibility. All integrated findings were initially assumed to be of high quality. Based on the assessment, the overall quality of evidence was then downgraded, as necessary, and categorized into four levels: high, medium, low, or extremely low. Table 3 shows the ConQual system scores for this review.

Table 3.

ConQual system scores and the specific reason

Synthesized findings	Type of research	Dependability	Credibility	ConQual score	Comments
Perceived Benefits	Qualitative	Remains unchanged*	Remains unchanged**	High	*Dependability: The findings came from 21papers remains unchanged as the majority of studies (15 out of 21) scored 4 or 5 on questions related to the appropriateness of the conduct of the study. **Credibility: As all findings are unequivocal and remain unchanged.
Perceived Shortcomings	Qualitative	Remains unchanged*	Remains unchanged**	High	*Dependability: The findings came from19 papers remains unchanged as the majority of studies (13 out of 19) scored 4 or 5 on questions related to the appropriateness of the conduct of the study. **Credibility: As all findings are unequivocal and remain unchanged.
Perceived Risks	Qualitative	Remains unchanged*	Remains unchanged**	High	*Dependability: The findings came from 14 papers remains unchanged as the majority of studies (10 out of 14) scored 4 or 5 on questions related to the appropriateness of the conduct of the study. **Credibility: As all findings are unequivocal and remain unchanged.
Perceived Needs	Qualitative	Remains unchanged*	Remains unchanged**	High	*Dependability: The findings came from 12 papers remains unchanged as the majority of studies (8 out of 12) scored 4 or 5 on questions related to the appropriateness of the conduct of the study. **Credibility: As all findings are unequivocal and remain unchanged.

Note: The process of evaluation is described in detail in Supplementary Material 2

Results

After careful screening of 764 initially identified articles and an additional 2 articles identified through manual searching, 24 studies comprising a total of 623 participants were ultimately included. The included studies originated from multiple countries, including China (n = 9), the United States (n = 4), Turkey (n = 3), Australia (n = 2), Thailand (n = 1), Spain (n = 1), Canada (n = 1), Malaysia (n = 1), Singapore (n = 1), and the United Arab Emirates (n = 1). Most studies were published in the past five years and were classified as original research. Through repeated review, the researchers identified 41 distinct findings, grouped them into 11 categories, and further synthesized these into four overarching summary findings.

Synthesized finding 1: Perceived benefits

This section presents the integrated findings on nursing students’ perceived benefits of using GAI tools. The findings were grouped into three categories: empowering experiences for self-directed learning, effortless experiences with tool use, and novel experiences of domain exploration.

Category 1: Empowering experiences for self-directed learning

Nursing students reported multifaceted skill development through GAI use. These tools offered rapid responses to students’ inquiries, enhancing learning efficiency. Some students noted that literature reviews could be completed within minutes. Others highlighted GAI’s ability to deliver personalized learning assistance. Furthermore, GAI’s extensive information resources were seen to broaden students’ knowledge and strengthen their academic confidence. Additionally, nursing students noted that GAI provided academic guidance, helped organize ideas, and improved language use, thus enhancing writing skills and overall academic quality.

It allows me to create work, reading lists, and summaries in minutes, rather than hours [22].

I receive individualized feedback and answers through using ChatGPT [37].

In the future, I am not afraid of being asked about obstetric vaccinations [25].

I think it can assist in academic writing, creating general outlines or refining language to strengthen arguments, making papers look more professional [35].

Category 2: Effortless experience with tool use

Nursing students considered GAI tools easy to use. Firstly, these tools featured user-friendly designs. Students described GAI interfaces as clear and intuitive, facilitating simple interactions, straightforward operations, and rapid responses. Secondly, tools offered free and flexible access. Students frequently mentioned the complimentary, round-the-clock availability of most GAI tools, alleviating financial barriers and improving usability.

It was simple to navigate between the options provided, and it was quick [24].

Because LLMs are free and there are no limits on usage, I can consult them as often as needed [20].

At any time, all I need is to open my laptop and communicate to ChatGPT [37].

Category 3: Novel experiences of domain exploration

Nursing students expressed considerable curiosity and interest in the emerging GAI field. Some students were interested in GAI’s core functions, multi-scenario capabilities, and specialized strengths across different tools. Additionally, students described exploring GAI’s potential capabilities as novel and engaging.

Providing fundamental nursing knowledge, translation services, PowerPoint outlines… and more; they are like a versatile toolbox [20].

They specialize in different areas: some are proficient in coding, others in diagramming, and still others in creating presentations [19].

It generated an anatomical image of the thyroid gland based on my prompts, and the image can be modified according to my inputs, making it more engaging than a traditional resource book [18].

Synthesized Finding 2: Perceived shortcomings

This section presents the integrated findings on nursing students’ perceived shortcomings when using GAI tools. The findings were grouped into four categories: concerns regarding information quality, limitations in artificial intelligence literacy, technical flaws, and resource constraints.

Category 4: Concerns regarding information quality

Nursing students expressed doubts about the quality of information generated by GAI. Several students reported significant concerns about the accuracy of the content provided. During use, students frequently encountered results that did not meet their expectations, leading them to question the information’s usefulness. Some students indicated that the absence of cited data sources reduced the credibility of the information. Additionally, responses were often perceived as overly generalized, vague, and lacking sufficient detail.

However, some [ChatGPT] models may provide incorrect answers [40].

The generated visual effects were perceived as inadequate, with outputs frequently being low quality or unrelated to intended content [41].

For instance, if you inquire about the most relevant literature on a theory, it will generate a list of references and provide DOI numbers; however, full-text retrieval is not possible [19].

ChatGPT produced long lists of interventions, which gave an impression that it answered my questions thoroughly. However, my own research found additional points that ChatGPT did not mention [21].

Category 5: Limitations in artificial intelligence literacy

Nursing students commonly exhibited low levels of artificial intelligence literacy. Some students reported excessive reliance on AI, which they perceived as diminishing their critical thinking abilities. Others were hesitant to use AI tools due to insufficient understanding of the technology and uncertainty regarding proper use. Additionally, nursing students experienced increased anxiety related to concerns about potential job displacement and future unemployment.

It reduced my critical thinking and problem-solving skills [37].

However, because it is new, we cannot easily trust it [23].

I’m not very skilled at using it [20].

AI feels like it might replace human work entirely [41].

Category 6: Technical flaws

Technical flaws in GAI caused negative experiences among nursing students. Some reported system instability issues, such as response failures during use. Other students indicated that the overly technical jargon used by GAI was difficult to understand, causing misunderstandings and increased cognitive burden. Additionally, students highlighted the lack of empathetic or supportive responses from GAI in emotionally complex situations.

I waited for a long time, but no answer came. I had to log out and log back in [27].

Sometimes it fails to grasp my meaning, forcing me to repeatedly adjust how I express myself [27].

When I typed ‘this question is too hard,’ the system didn’t recognize my frustration and just repeated the question. It felt like arguing with a machine [36].

Category 7: Resource constraints

Nursing students encountered resource and geographical limitations when using GAI tools. Some students reported difficulties accessing many GAI tools due to high licensing fees and geographic restrictions. Additionally, many indicated a lack of formal training in GAI usage. Some students also noted institutional restrictions regarding GAI use.

Licensed tools require payment for usage rights, which seems expensive, and students don’t have the income for this [19].

We can’t access certain large language models through official channels in our country [20].

Formal training on AI tools? None [38].

The school prohibits us from using large language models [20].

Synthesized finding 3: Perceived risks

This section presents the integrated findings on nursing students’ perceived risks of using GAI tools. The findings were grouped into two categories: divergent perceptions regarding data security and ethical and moral challenges.

Category 8: Divergent perceptions regarding data security

Nursing students displayed varying attitudes toward data security when using GAI. Some students expressed significant concerns about personal privacy protection. Students using GAI for writing support were particularly worried about potential leaks of thesis-related information. However, other students showed limited awareness of data security, adopting either permissive or indifferent attitudes toward privacy breaches.

With the increasing risk of privacy breaches, I’m worried about information security [27].

After all, I haven’t submitted the paper yet. I’m concerned that my answers might be directly provided to someone else if they ask the same question [19].

Information currently seems secure enough, and with legal protections, I still generally trust it [27].

I don’t think this constitutes a data breach, and there’s no benefit in disclosing my information [19].

Category 9: Ethical and moral challenges

Nursing students encountered ethical dilemmas related to GAI use. Some students acknowledged receiving academic writing assistance from GAI but expressed concerns about compromising academic integrity. Conversely, other students avoided GAI, viewing its use as a form of “cheating” that undermines fairness in education. Additionally, nursing students raised concerns about privacy and ethical issues associated with GAI’s handling of patient information.

If an article is entirely AI-generated, it could harm originality and lead to misconduct [41].

Fairness is compromised if students use GAI to complete and submit assignments [22].

I’m not sure whether using GenAI with patient data violates privacy [41].

Synthesized finding 4: Perceived needs

This section presents the integrated findings on nursing students’ perceived needs for using GAI tools. The findings were grouped into three categories: desire for policy and educational support, and expectations for developing and optimizing professional tools.

Category 10: Desire for policy and educational support

From a policy standpoint, nursing students expressed the need for stronger regulatory frameworks to ensure appropriate GAI use. Specifically, some students requested clear guidelines defining ethical standards governing GAI usage.

I think the school should provide clear guidelines. Sometimes I’m unsure about what constitutes plagiarism, making it confusing [40].

Regarding educational support, nursing students emphasized the need for technical training, ethics education, and interdisciplinary collaboration. Some students reported limited understanding of GAI technologies, hindering their effective use of GAI-based learning tools. Consequently, they suggested integrating relevant training courses into the curriculum. Others highlighted the importance of formally embedding GAI into nursing education. Additionally, a significant number advocated education on ethical boundaries.

I hope schools provide lectures or courses on proper GAI usage [19].

I encourage our school curriculum to integrate with ChatGPT [23].

It seems necessary to educate students clearly about ethics, rules, boundaries, appropriate usage, and recognizing right from wrong [35].

Category 11: Expectations for developing and optimizing professional tools

Technological optimization emerged as a critical form of support required by nursing students. Some students reported challenges with low-quality information generated by GAI, attributing this to insufficient professional knowledge and overly general or vague responses. They therefore suggested developing domain-specific GAI tools tailored to medical and nursing contexts. Additionally, students recommended that GAI systems incorporate accessibility-focused features, such as color adjustments and voice-guided navigation, to better assist nursing students with disabilities.

It should be trained using existing medical and nursing expertise or high-quality literature to create a specialized medical chatbot [18].

Because I’m color blind and must interact with chatbots featuring contrasting fonts and background colors, it’s challenging to read and navigate the available options [24].

Discussion

This study synthesized qualitative evidence on nursing students’ experiences with GAI, including 24 studies, and identified four main themes: perceived benefits, perceived shortcomings, perceived risks, and perceived needs. Notably, the sample sizes of the included studies varied considerably, ranging from 5 to 114 participants. Three studies had extreme sample sizes (5, 99, and 114 participants) [24, 25, 28], whereas the remaining 21 studies had more concentrated sample sizes, ranging from 11 to 48 participants. Nevertheless, the findings from these three studies did not generate new themes; instead, they corroborated those identified in the other studies. This indicates that variations in sample size did not substantially influence the overall findings.

We found that the four synthesized findings of this study corresponded to some extent with the Unified Theory of Acceptance and Use of Technology (UTAUT), as shown in Fig. 2. The UTAUT model suggests that an individual’s adoption of new technology is primarily influenced by four core constructs: performance expectancy (perceived usefulness), effort expectancy (perceived ease of use), social influence (perceived social pressure), and facilitating conditions (perceived external support) [42]. The primary aim of this study was to promote the appropriate and effective use of GAI among nursing students. Accordingly, an in-depth examination was conducted on the factors underlying nursing students’ diverse experiences with GAI, applying the UTAUT model to explain their technology adoption behaviors and identify effective educational strategies.

Fig. 2.

Mapping of major findings to UTAUT constructs

Maintaining and enhancing positive experiences can encourage nursing students’ willingness to use GAI. Our research indicated that nursing students perceive GAI tools as empowering for learning, encompassing increased efficiency, broader knowledge acquisition, personalized learning, and academic writing support. This finding demonstrates that nursing students have high levels of performance expectancy regarding GAI. This is because GAI possesses advanced capabilities in data storage, information processing, and content generation, allowing rapid generation of responses tailored to user needs [43]. Additionally, nursing students indicated a high effort expectancy when using GAI, perceiving it as highly responsive and user-friendly. This perception aligns with GAI’s intuitive interface, which requires no complex commands and allows interaction through everyday language [44]. Nursing students also expressed feelings of novelty and enjoyment during interactions with GAI tools. These experiences originate from the tools’ diverse functionalities: beyond basic information processing, GAI tools can generate images and even produce PowerPoint presentations. Such capabilities exceed students’ initial expectations, thereby stimulating exploratory curiosity [19]. According to the UTAUT model, performance expectancy and effort expectancy significantly influence nursing students’ attitudes and intentions toward using GAI. Therefore, educators can enhance students’ technical skills through structured GAI-related courses or training programs [44]. Additionally, establishing dynamic incentive mechanisms may further stimulate students’ motivation and initiative in applying GAI tools [45], reinforcing their positive experiences.

Negative experiences hinder nursing students’ adoption and use of GAI tools and therefore require careful academic attention. At the technical level, nursing students were dissatisfied with the quality of GAI-generated information, which was frequently overly general, lacked proper source attribution, or was inaccurate. Such negative experiences consequently reduced students’ performance expectancy. This issue arises because GAI systems may produce “hallucinations”, generating plausible yet incorrect information to address gaps when confronted with insufficient or unclear training data [46]. Technical limitations inherent in GAI tools, such as system instability and inadequate emotional responsiveness, further contributed to negative user experiences. These discrepancies reflect a mismatch between nursing students’ expectations of GAI and its actual capabilities [36]. Secondly, at the individual level, this study found that nursing students faced challenges related to insufficient AI literacy, including limited understanding, inadequate operational skills, over-reliance on these tools, and concerns about being replaced when using GAI tools. These issues increase the burden of use and weaken effort expectancy. This may be attributed to the current lack of emphasis on AI literacy education within nursing education programs [47]. Finally, at the environmental level, this study found that some nursing students experienced limited access to resources due to multiple barriers, including cost, geographic constraints, and institutional restrictions, which indicate severely constrained facilitating conditions for the use of GAI tools. This reflects the current lack of support for GAI technology within the nursing education system [28]. From a technology acceptance perspective, negative experiences significantly undermine nursing students’ performance expectancy, effort expectancy, and indicate insufficient facilitating conditions. Therefore, educational institutions should update pedagogical approaches, provide institutional technological support, and encourage nursing students’ effective GAI use [4]. Educators should implement structured, tiered AI literacy curricula covering fundamental concepts, interaction skills, and information quality evaluation to enhance nursing students’ technological competence and critical thinking abilities [48].

Effectively managing risks related to data privacy, security, legal compliance, and ethical considerations is essential for nursing students to responsibly and effectively use GAI. The integration of GAI into nursing education and clinical training has become an inevitable trend. However, its use inevitably involves personal information, learning records, and patient data, thus posing potential risks of data leakage. Our research indicated that some nursing students had misunderstandings about privacy breaches, either failing to recognize potential information leakage associated with GAI or adopting indifferent attitudes toward such risks. These misconceptions warrant serious attention and targeted educational intervention. Furthermore, increasing ethical concerns pose substantial challenges to the effective use of GAI among nursing students. Research showed that nursing students worried that using GAI for learning, academic writing, and clinical decision-making might compromise academic integrity, undermine fairness in learning environments, and blur legal accountability. These concerns arise from the current lack of clear ethical guidelines and systematic education on ethical boundaries [32]. Therefore, educational institutions should establish clear ethical guidelines for GAI use to prevent overreliance and clarify accountability [49]. Educators should implement targeted training on data security and AI ethical standards to enhance students’ awareness of privacy and legal responsibilities [50]. Notably, these findings highlight limitations of the UTAUT model in addressing the complex experiences of nursing students using GAI. Specifically, the model provides limited guidance regarding ethical considerations; thus, findings from this study could inform extensions of the model.

Addressing nursing students’ needs is fundamental to promoting effective GAI use. The UTAUT model indicates that facilitating conditions significantly influence nursing students’ use of GAI, with students’ needs forming a crucial component of these conditions. Research demonstrated that nursing students’ needs primarily involved two areas: resource optimization and technical optimization. The need for policy and educational resources has been discussed earlier, particularly in relation to mitigating negative experiences and addressing ethical risks. Therefore, this discussion focuses on technical optimization needs. Our research indicated that the development of specialized tools for medical fields and ongoing technical enhancements aligned with many nursing students’ expectations. This expectation arises from limitations in general-purpose GAI tools, which often demonstrate inadequate accuracy in specialized knowledge, overly generalized responses, and insufficient accessibility features [21]. Therefore, technology development teams should prioritize user needs through in-depth research, develop dedicated GAI tools tailored to nursing, and emphasize accessible interface designs to broaden the potential user base.

Recommendations

This study proposes 7 recommendations at both the practice and policy levels. The detailed recommendations are summarized in Table 4.

Table 4.

Practice/policy recommendations

Recommendations	Supporting themes	Grade
R1: Educators should integrate GAI training into nursing curricula, emphasizing students’ ability to effectively apply GAI tools in academic and clinical contexts.	Category 1: Empowering experiences for self-directed learning Category 3: Novel experiences of domain exploration Category 5: Limitations in artificial intelligence literacy Category 10: Desire for policy and educational support	B
R2: Educators should implement tiered AI literacy curricula, covering foundational concepts, interaction skills, and information quality evaluation, to enhance nursing students’ technological skills and critical thinking.	Category 4: Concerns regarding information quality Category 5: Limitations in artificial intelligence literacy Category 8: Divergent perceptions regarding data security	B
R3: Educators should include instruction on data security and AI ethical standards within nursing curricula, guiding students toward an accurate understanding and effective use of AI, while improving their awareness of privacy and legal responsibilities.	Category 8: Divergent perceptions regarding data security Category 9: Ethical andmoral challenges Category 10: Desire for policy and educational support	B
R4: Technology developers should adopt a user-centered approach, focusing on the development of domain-specific GAI tools tailored for nursing students to provide targeted support for professional learning and clinical knowledge acquisition.	Category 4: Concerns regarding information quality Category 6: Technical flaws Category 11: Expectations for developing and optimizing professional tools	B
R5: Technology developers should regularly update interfaces and functions, prioritize accessibility and user-centered design, reduce usage barriers, and improve GAI tool accessibility for students from diverse backgrounds.	Category 2: Effortless experience with tool use Category 11: Expectations for developing and optimizing professional tools	B
R6: Educational institutions should update educational philosophies, secure institutional access to GAI tools, and encourage their appropriate use among nursing students, thus creating a supportive environment for effective GAI utilization.	Category 7: Resource constraints Category 10: Desire for policy and educational support	B
R7: Educational institutions and hospitals should establish clear ethical guidelines for GAI use, explicitly addressing academic integrity, defining responsibilities, and promoting standardized GAI practices among nursing students.	Category 9: Ethical and moral challenges Category 10: Desire for policy and educational support	B

Note: Recommendations 1–5 represent practice-level implications, and recommendations 6–7 represent policy-level implications. For all recommendations, the population-phenomenon-context is defined as nursing students-GAI experience-learning/clinical context

Limitations

This meta-synthesis has several limitations. (1) Only literature published in Chinese and English from ten countries was included, potentially excluding relevant studies in other languages and introducing geographical bias. (2) Gray literature (i.e., scholarly materials such as dissertations, conference abstracts, and institutional reports, not formally published or commercially distributed) was not included, possibly leading to incomplete coverage of relevant research. (3) The methodologies used in the included studies were limited, primarily employing descriptive designs and semi-structured interviews. (4) Quality appraisal indicated that some studies received ratings of “unclear” or “no” on specific criteria, potentially impacting result interpretation. (5) As this qualitative meta-synthesis relied on secondary data, the absence of original context and potential interpretive bias may have influenced the synthesized results. (6) The findings primarily reflect the shared experiences of nursing students using GAI, while potential variations related to participant characteristics, GAI tool types, and usage scenarios were not extensively examined.

Conclusion

This study explored nursing students’ experiences with GAI, identifying four synthesized findings: perceived benefits, perceived shortcomings, perceived risks, and perceived needs. Nursing students’ experiences with GAI were found to be complex and multifaceted. Although students derived substantial benefits from GAI, they also faced limitations related to technological constraints, individual factors, and environmental conditions. Additionally, concerns regarding data security and ethics were increasingly significant. Therefore, educational institutions, educators, and technology developers should prioritize addressing nursing students’ needs by enhancing positive experiences, mitigating negative experiences, and effectively managing ethical risks. This necessitates the strengthening of technical, educational, and policy support frameworks to ensure the appropriate, safe, and efficient use of GAI tools among nursing students.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Sihua Wang: Writing-original draft, Writing-Review & Editing, Methodology, Investigation, Formal analysis, Data curation. Shuzhen Niu: Writing-original draft, Writing-Review & Editing, Methodology, Investigation, Formal analysis, Data curation. Xinyu Zhang: Software, Investigation, Methodology, Data curation. Qianqian Cao: Software, Investigation, Methodology, Validation.Qihong Li: Software, Resources, Methodology, Validation.Jia Yu: Software, Resources, Methodology, Visualization.Jinxia Jiang: Writing-Review & Editing, Project administration, Funding acquisition, Conceptualization, Supervision. Li Zeng: Writing-Review & Editing, Project administration, Funding acquisition, Conceptualization, Supervision.

Funding

This review was supported by the Youth Clinical Research Project of Tongji Hospital, Tongji University [ITJ(QN)2207], the Interdisciplinary Research Project of Tongji University [2026-0313-YB-13], and the 2025 Shanghai Philosophy and Social Sciences Planning Project [2025BJC005].

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.