AI‐Based Learning Experiences of Nursing Students: A Qualitative Study

ABSTRACT

Background

Global nursing courses are rapidly attempting to utilise large language models (LLMs), and generative AI such as ChatGPT now traverses the traditional classroom–clinical divide.

Methods

This study conducted a qualitative descriptive investigation to explore nursing students' specific experiences and transformations in AI‐based learning.

Results

Four interrelated themes were identified: AI‐Driven Transformation of Nursing Students' Learning Behaviour Patterns; Emotional‐Cognitive Shifts in Nursing Students Through AI‐Based Learning; Self‐Directed Learning Outcomes in Nursing Students Using AI‐Based Learning and Adaptive Challenges.

Conclusions

The findings offer evidence‐based insights into how nursing students can leverage AI to navigate rapidly changing nursing education and healthcare environments worldwide.

1. Introduction

Large language models (LLMs), epitomised by the release of ChatGPT in late 2022, have precipitated a paradigmatic shift in technology‐enhanced education [1, 2]. AI‐based learning systems built upon these generative models are attracting unprecedented attention within the medical and nursing sciences [3, 4]. Accumulating evidence indicates that artificial intelligence is rapidly evolving from a supportive tool into a transformative driver capable of reconfiguring the epistemological, pedagogical, and practical foundations of medical education [5]. Simultaneously, AI‐based learning has emerged as a significant learning modality, encompassing medical classroom instruction [6], clinical practice, administrative support, and other medical education domains.

Applications of AI in nursing encompass teaching evaluation, the creation of clinical simulation cases [7], and the provision of personalised learning support [8], with the goal of enhancing both educational experiences and outcomes. The integration of artificial intelligence optimises the learning experience for nursing students, enhances their adaptability to clinical environments, and prepares them for the challenges posed by the rapidly evolving healthcare landscape [9]. Recent systematic reviews confirm AI's beneficial effects on nursing students' knowledge acquisition, skills development, perceptions and attitudes. Additionally, numerous studies have examined issues stemming from AI tool usage in nursing education, including moral and ethical concerns [10, 11].

Current research predominantly emphasises the functional development and application efficacy of AI tools [12], while systematic learner‐centred exploration of how students deal with AI‐generated knowledge during the learning process is still insufficient. Theoretically, the integration of AI into education can be examined through the lens of Zimmerman's Self‐Regulated Learning (SRL) theory [13, 14] and Bandura's Social Cognitive Theory (SCT) [15, 16]. SRL defines learning as a proactive cycle of forethought, performance control, and self‐reflection; SCT stresses self‐efficacy and environmental feedback in shaping cognitive and emotional behaviours. For nursing students balancing clinical internships and academic exams, AI tools boost self‐efficacy via instant feedback, yet may disrupt the SRL cycle by weakening self‐reflection. Notably, cognitive‐emotional experiences and autonomous learning of nursing students in AI‐enabled learning contexts remain under‐theorised.

This qualitative study explored their concrete AI learning experiences, including transformations in learning behaviour patterns, emotional‐cognitive shifts, and self‐directed learning outcomes, to inform AI learning strategy optimisation and enhance their adaptability in evolving healthcare environments.

2. Methods

2.1. Study Design

This study adheres to the consolidated criteria for reporting qualitative research (COREQ) [17]. A qualitative approach was adopted in this research to explore the experience of nursing students through AI‐based learning. Qualitative research is appropriate for capturing authentic experiences and obtaining a profound comprehension of participants' viewpoints. It is particularly beneficial for uncovering emotional associations and transformations of self‐directed learning ability that are experienced by nursing students in an AI‐based learning environment.

2.2. Setting and Participants, and Study Context

The study took place within a nursing college at a medical university, which provides a comprehensive nursing curriculum. Importantly, during the period when this study was conducted, this college lacked formal guidelines and explicit policies governing the use of generative AI in coursework. Furthermore, AI use was not formally integrated into the nursing curriculum, nor did students receive systematic training on AI tools from the faculty. Consequently, participants' use of AI was entirely self‐initiated, informal, and self‐directed.

Participants were recruited using purposive sampling. The inclusion criteria were: (1) senior undergraduate or first‐year graduate nursing students; (2) having completed at least 11 months of clinical hospital internship; and (3) having actively used generative AI tools for learning purposes during the previous semester. Students who had never used AI tools for learning were excluded. Potential participants were approached by the primary researcher through course announcement groups or WeChat groups. A total of 27 students were invited; 2 declined due to scheduling conflicts, resulting in a final sample size of 25 participants. Sample size was determined by data saturation, assessed iteratively during concurrent data collection and analysis. Following analysis of 23 interviews, the core thematic framework had stabilised, and the additional 2 interviews yielded no new codes or meaning units, confirming that meaning saturation was achieved.

2.3. Data Collection

Interviews were conducted between July 2025 and September 2025. Semi‐structured, face‐to‐face interviews were employed to facilitate a comprehensive exploration of the students' AI‐based learning experiences [18]. Primary interview dialogues centred on usage variations: participants were asked to describe the specific AI tools used (e.g., Large Language Models such as DeepSeek and Doubao), their usage frequency, and learning contexts (e.g., examination preparation, internship reporting, or systematic study planning). Additionally, open‐ended questions explored transformations in learning behaviour patterns, emotional‐cognitive shifts, self‐directed learning outcomes, and adaptive challenges. Each interview session lasted approximately 20–40 min, allowing participants sufficient time for reflection. All interviews were audio‐recorded with consent and transcribed verbatim within 48 h for subsequent analysis.

2.4. Data Analysis

All interview transcripts were anonymised and imported into Microsoft Word for data management. The analysis was based on thematic analysis. Data were analysed using the six‐step thematic analysis framework proposed by Braun and Clarke (2006) [19] This involved: (1) familiarisation with the data through repeated reading; (2) generating initial codes systematically across the dataset; (3) collating codes into potential themes; (4) reviewing themes against the coded extracts; (5) defining, refining and naming themes; and (6) producing the final report.

To ensure analytical rigour, two researchers (J.M. and X.Y.P.) independently reviewed the transcripts and generated initial codes. Following the independent coding phase, the two researchers met to compare their initial coding results. Discrepancies in code definitions or thematic categorisation were resolved through detailed discussion. When consensus could not be reached, a third senior researcher (W.D.) intervened as an adjudicator to facilitate a collaborative resolution, thereby ensuring the robustness and transparency of the final thematic framework.

2.5. Rigour and Reflexivity

Methodological rigour was maintained using Lincoln and Guba's criteria: credibility, transferability, dependability, and confirmability [20]. Credibility was enhanced through member checking, where detailed synthesised documents of the interview findings were provided to 23 participants to verify the accuracy of the researchers' interpretations. Transferability was established by providing thick descriptions of the participants' backgrounds and the informal AI learning context. Dependability and confirmability were achieved through peer debriefing with other members of the research team and by maintaining a detailed audit trail that documented all methodological decisions and coding iterations.

Regarding researcher reflexivity, the interviews were conducted by the first and second authors (J.M. and X.Y.P.), who are nursing graduate students with formal training in qualitative research methods. The researchers acknowledged their pre‐existing assumptions regarding the potential benefits and risks of AI in nursing education and engaged in bracketing to prevent these views from influencing the data collection. Furthermore, conducting interviews as peers (graduate students) rather than faculty members helped minimise hierarchical power dynamics, encouraging participants to share their authentic experiences more freely. Nevertheless, participants were explicitly assured that their responses would remain completely confidential and would not be shared with course instructors. This reflexive stance helped maintain analytical neutrality throughout the study.

2.6. Ethical Considerations

The research project received approval from the relevant Committee (identification code: 202455). Research protocols consistently observed the Declaration of Helsinki's ethical standards. Prior to the interview, verbal consent acquisition was obtained from all participants. Participants were thoroughly apprised of the study's purpose, their right to withdraw at any time without penalty, and the strict confidentiality of their data. During the focus interviews, the research personnel monitored the participants' involvement and recorded their realistic feelings.

3. Results

3.1. Participants Characteristics

A total of 23 (N = 23) participants were interviewed and data were analysed. All participants' demographic characteristics are presented in Table 1. Twenty‐two participants (n = 22) were female, and one participant (n = 1) was male. The participants (n = 23) were aged between 20 and 22 years. Among the participants, twenty (n = 20) held bachelor's degrees in nursing, and three (n = 3) were admitted to nursing graduate programmes. The nursing students offered insights into AI‐based learning experiences. Informed consent was acquired from every participant, who voluntarily disclosed their perspectives and feelings of learning experiences to the researchers.

Table 1.

Participant characteristics (n = 23).

Participant	Gender	Academic qualification	Most commonly used AI tools	Frequency	Primary learning context/purpose
F1	Female	BSN	DeepSeek	NR	Clinical case reasoning & Drafting nursing care plans
F2	Female	BSN	Quark Browser/DeepSeek	NR	Concept explanation & Reviewing clinical procedures
F3	Female	BSN	ChatGPT/Grammarly	Twice a month/Once a week	Coursework structuring & Systematic study planning
F4	Female	MSN	Doubao	Twice a week	Internship reporting & Dynamic schedule management
F5	Female	BSN	DeepSeek/Doubao	Almost daily	Literature structural extraction & Essay polishing
F6	Male	MSN	DeepSeek/Doubao	Almost daily	Exam preparation & Disease knowledge retrieval
F7	Female	BSN	DeepSeek/Doubao	NR	Exam preparation & Literature retrieval
F8	Female	BSN	ChatGPT/DeepSeek	NR	Exam preparation & Clinical case reasoning
F9	Female	BSN	DeepSeek/Quark Browser	NR	Literature structural extraction & Essay polishing
F10	Female	BSN	DeepSeek/Doubao	Once a week	Exam preparation & Systematic study planning
F11	Female	BSN	DeepSeek/Kimi	Two to three times per week	Essay polishing & Systematic study planning
F12	Female	BSN	DeepSeek/Doubao	Twice a week	Literature retrieval & Essay polishing
F13	Female	BSN	ChatGPT/DeepSeek	NR	Internship reporting & Academic text optimisation
F14	Female	BSN	Ernie Bot	Three times per week	Exam preparation & Internship reporting
F15	Female	BSN	DeepSeek/Doubao	Two to three times per week	Internship reporting & Research brainstorming
F16	Female	BSN	Kimi/DeepSeek	Two to three times per week	Literature retrieval & Thesis defence structuring
F17	Female	MSN	Kimi/Claude	Almost daily	Literature retrieval& Essay polishing
F18	Female	BSN	ChatGPT	NR	Academic reporting & Essay polishing
F19	Female	BSN	DeepSeek	NR	Exam preparation & Research brainstorming
F20	Female	BSN	Doubao	NR	Exam preparation & Disease knowledge retrieval
F21	Female	BSN	ChatGPT/Metawrite	Almost daily	Literature retrieval & Research brainstorming
F22	Female	BSN	Doubao/Quark Browser	NR	Coursework structuring & Systematic study planning
F23	Female	BSN	Doubao	NR	Concept explanation & Reviewing clinical procedures

Abbreviations: BSN, bachelor of science in nursing; MSN, master of science in nursing; NR, not reported.

3.2. Thematic Overview of Qualitative Data (Interviews)

The four primary themes emerging from analytical procedures are outlined in Figure 1. Four themes were identified to explore Nursing Students' AI‐based learning experience of behaviour patterns, emotional‐cognitive shifts, self‐directed learning outcomes, and adaptive challenges encountered.

Figure 1.

Four main themes.

To ensure transparency and clear alignment between themes, subthemes, and supporting data, Table 2 provides a comprehensive summary of the coding framework along with representative verbatim extracts.

Table 2.

Synthesis results framework.

Themes	Sub‐Themes	Codes	Illustrative quotes
AI‐Driven Transformation of Nursing Students' Learning Behaviour Patterns	Autonomous learning reinforcement	(1) Learning motivation	‘The answers from AI are definitely more comprehensive…its logical thinking is very rigorous, linking one step to another, which helps me improve my own logical rigour, that is, my way of thinking.’ (F1)
			‘Because I think AI brings great convenience, allowing you to solve problems immediately. It prevents you from procrastinating…but the immediacy of solving problems now can also enhance learning enthusiasm to a certain extent.’ (F22)
		(2) Individualised needs	‘The information provided by AI is sometimes too general and lacks pertinence. I hope it can be more tailored to personal learning situations.’ (F10)
	Construction of personalised learning paths		‘I hope AI can optimise answers according to my question‐asking habits and knowledge level, rather than being uniform.’ (F22)
		(2) Align strategy	‘Adjust the learning goals generated by AI according to personal knowledge level, focusing on learning weak links and reducing time on easy content’ (F3)
			‘If possible, I actually hope that AI can be combined with VR in the future. Perhaps some patients are not very willing to cooperate with us, but if we use VR…to have such feedback, I feel that this statement will slightly reduce the sense of distrust they have towards us nurses compared to when….’ (F23)
Emotional‐Cognitive Shifts in Nursing Students Through AI‐Based Learning	The emotional dependence and connection of nursing students towards AI tools	(1) Learning interest and confidence	‘When I couldn't figure out a clue, I asked the AI. It would provide me with many solutions to this problem, allowing me to select from them and deeply think about the most suitable approach and choice… this strengthened my confidence in learning.‘ (F5)
		(2) Factors affecting choice	‘The answers he provides are more accurate and closer to real life, which makes me more inclined to choose it. Another reason is its human‐like quality…If AI could, um, get into people's hearts and have better empathy with humans, I think this is also a reason for choosing it’ (F11)
	Emotional regulation of nursing students based on the learning process of AI	(1) Coping with Setbacks	‘“AI provides a more comprehensive and multi‐layered analysis. It presents information in a clear and understandable manner, rather than combining overly complex and convoluted sentences.’ (F8)
		(2) Dependence	‘“But I think to a certain extent, if I rely on it too much, I will also lose some time for my own self‐reflection. One would overly rely on this kind of technology, and sometimes they wouldn't have their own thoughts. I think this is also somewhat harmful.” (F22) “It could diminish my problem‐solving abilities because I've become reliant on AI technology. If I were to suddenly detach from AI technology, my ability to solve problems independently might decline.’ (F19)
Self‐Directed Learning Outcomes in Nursing Students Using AI‐Based Learning	Information literacy and self‐management ability	(1) Evaluation	‘In fact, the information given by AI is very long and comprehensive, so I must extract from it. That is, the content related to the disease I am dealing with, the teacher's question, and the project I am working on. I cannot say it answers the parts I don't want or the parts I already know. I must filter out the parts I don't know’ (F8)
	Critical and innovative thinking ability	(2) Acquisition	‘Retrieve relevant information on clinical issues through keyword matching to ensure content accuracy.’ (F6)
			‘Generally, AI‐provided information includes main titles and subheadings. When I filter information, I usually first scan through the directions and titles it lists, and then focus on the content relevant to me.’ (F10)
		(1) Reflection	‘When I have my own thoughts, I won't deliberately use AI. I only rely on it when my thoughts are confused. I don't want it to replace human spontaneous thinking’ (F1)
			‘AI is highly effective as a tool for information extraction and integration, saving time on initial preparations and allowing you to focus on tasks that require your own thinking. However, it's advisable not to rely on it too heavily. You should have the ability to critically assess the feedback it provides. Given the current limitations of AI development, it cannot completely replace your work.’ (F5)
		(2) Inspiration	‘The AI image generation function is quite interesting. For instance, if I read a book and ask it to generate pictures based on the scenery in the book, it can help me better understand the text and also bring me new ideas.’ (F11)
			‘For example, when using Meta‐analysis, in the topic selection stage, it will preliminarily screen according to specific requirements and relevant journal content. At that time, I had no clear direction, so I could hand over this problem to AI to see which aspects were more innovative.’ (F19)
	Perceived readiness for clinical decision‐making	(1) Proficiency	‘This also fosters a sense of proactive engagement. For instance, regarding the proficiency of certain procedures, AI can provide feedback and correct me on aspects where I may not have been thorough or precise. It doesn't merely follow standardised approaches…For my upcoming clinical practice, this actually provides me with a highly structured and standardised template. I can …and feel more confident in my future clinical practice’ (F23)
Adaptive Challenges	Delayed iterative updating of AI knowledge bases	(1) Delayed AI update	‘In the case of intravenous infusion, it may provide a very detailed and comprehensive process. However, in clinical practice, you need to further assess the patient's condition, such as whether the patient is obese, emaciated, oedematous, or has other complex situations. AI can provide patients with some basic information and educational content in outpatient and inpatient departments, but clinical decisions still need to be made by medical staff based on the actual situation.’ F21)
	Issues related to efficiency and learning depth	(1) Fragmented knowledge acquisition	‘It might make me lazy and inclined to cut corners, as if having AI means I can skip any important steps… You need to think critically about the feedback AI provides and refine it further to truly grasp the overall understanding of the task’ (F5)

3.3. Theme 1: AI‐Driven Transformation of Nursing Students' Learning Behaviour Patterns

In an AI‐based learning environment, nursing students exhibited profound transformations in their learning behavioural patterns. Rather than merely using AI as a search engine, students actively integrated tools like DeepSeek and Doubao as cognitive scaffolds to streamline complex academic and clinical tasks.

3.3.1. Subtheme 1: Autonomous Learning Reinforcement

Nearly every participant mentioned AI‐based learning as an emerging supplementary educational tool in the current phase. Compared with traditional learning methods, it is more efficient and comprehensive, significantly improving nursing students' learning effectiveness. It helps foster good study habits and enhances learning proactivity, particularly among students who lack planning or tend to procrastinate.

The answers from AI are definitely more comprehensive. Sometimes I wonder why I didn't think of that, which might motivate me…it's logical thinking is very rigorous, linking one step to another, which helps me improve my own logical rigour, that is, my way of thinking.

(F1)

Uh, because I'm not really someone who likes to make plans. So, in traditional learning, if I'm required to make a plan…when reviewing, it can get quite chaotic. But with AI tools, they help me generate it automatically. Well, although I might need to make some adjustments during implementation, the overall framework is fine

(F3)

Previously, if problems couldn't be resolved promptly or if a solution approach wasn't readily available, they would often be deferred. Consequently, the positive feedback mechanism by AI‐driven subtly encourages nursing students to address a greater number of issues actively.

Because I think AI brings great convenience, allowing you to solve problems immediately. It prevents you from procrastinating. Sometimes, due to procrastination, problems are delayed. But the immediacy of solving problems now can also enhance learning enthusiasm to a certain extent.

(F22)

3.3.2. Subtheme 2: Construction of Personalized Learning Paths

Despite the benefits, participants expressed a critical need for deeper individualisation.

The information provided by AI is sometimes too general and lacks pertinence. I hope it can be more tailored to personal learning situations.

(F10)

I hope AI can optimise answers according to my question‐asking habits and knowledge level, rather than being uniform.

(F22)

AI‐based learning has emerged as a contemporary trend in both nursing classroom education and clinical decision‐making. Students' expectations have shifted from generic information retrieval to context‐aware academic mentorship. Specifically, they seek AI tools that can deliver customised learning plans aligned with their individual trajectories and knowledge proficiencies to enhance self‐directed learning. Given their limited clinical decision‐making experience, students also strongly desire AI‐powered scenario simulations and case‐based practice exercises, such as combining AI with VR for puncture simulations, to safely practice clinical reasoning without compromising patient safety and build a solid foundation for future independent clinical practice.

Adjust the learning goals generated by AI according to personal knowledge level, focusing on learning weak links and reducing time on easy content

(F3)

If possible, I actually hope that AI can be combined with VR in the future, for instance, during our operation process. Perhaps some patients are not very willing to cooperate with us, but if we use VR, for instance, during a puncture, we can provide such a virtual scene…to have such feedback, I feel that this statement will slightly reduce the sense of distrust they have towards us nurses compared to when we were operating on some real people.

(F23)

3.4. Theme 2: Emotional‐Cognitive Shifts in Nursing Students Through AI‐Based Learning

During the practical application of AI‐based learning, nursing students' attitudes and emotional perceptions undergo subtle yet significant transformations. This primarily involves changes in their acceptance of AI tools, their confidence in learning, and their emotional regulation throughout the learning process.

3.4.1. Subtheme 1: The Emotional Dependence and Connection of Nursing Students Towards AI Tools

AI interaction elicited a complex duality of emotions: enhanced academic self‐efficacy coupled with emerging psychological dependence. Several nursing students reported that AI‐based learning significantly bolstered their learning confidence and problem‐solving abilities, which undoubtedly represents a positive psychological boost.

When I couldn't figure out a clue, I asked the AI. It would provide me with many solutions to this problem, allowing me to select from them and deeply think about the most suitable approach and choice for the situation I was encountering at the moment. This strengthened my confidence in learning.

(F5)

Several nursing students favoured widely‐used AI tools, such as DeepSeek and Meta‐AI, due to their responses being highly attuned to their learning requirements, offering prompt guidance and assistance.

The answers he provides are more accurate and closer to real life, which makes me more inclined to choose it. Another reason is its human‐like quality…If AI could, um, get into people's hearts and have better empathy with humans, I think this is also a reason for choosing it

(F11)

I think it's a matter of usage habits and whether the feedback from the AI software is reasonable. For example, with DeepSeek, Doubao, and ChatGPT, I might find it more convenient to use DeepSeek, or maybe the first few times I used Doubao, the feedback wasn't very satisfactory, so I would continue to choose DeepSeek to help me solve problems because it meets my expectations to a greater extent in various ways.

(F7)

Several participants noted that they adopted AI‐based learning tools mainly due to user‐friendliness, attractive interface design and positive media reviews. However, this emotional connection also induced resistance when AI hallucinated or provided false references, shattering the illusion of AI infallibility.

Well, each AI has different app icons, and I prefer those with a more streamlined appearance. Additionally, I look for ones that are convenient to use and have a simplified design. Another factor is that I tend to rely on reviews from video assessments and the feedback provided… If it provides false or non‐existent references, I would refuse to use AI for this aspect of referencing.

(F13)

3.4.2. Subtheme 2: Emotional Regulation of Nursing Students Based on the Learning Process of AI

Nursing students had diverse emotional experiences when using AI‐based learning tools. The logically organised content provided by these tools relieved their apprehension of challenging tasks and facilitated problem‐solving.

AI provides a more comprehensive and multi‐layered analysis. It presents information in a clear and understandable manner, rather than combining overly complex and convoluted sentences.

(F8)

Nevertheless, extended reliance on the convenience afforded by AI could potentially cultivate dependency and hinder the development of independent thinking.

But I think to a certain extent, if I rely on it too much, I will also lose some time for my own self‐reflection. One would overly rely on this kind of technology, and sometimes they wouldn't have their own thoughts. I think this is also somewhat harmful.

(F22)

It could diminish my problem‐solving abilities because I've become reliant on AI technology. If I were to suddenly detach from AI technology, my ability to solve problems independently might decline.

(F19)

3.5. Theme 3: Self‐Directed Learning Outcomes in Nursing Students Using AI‐Based Learning

In the AI‐based learning environment, the changes in nursing students' self‐directed learning outcomes serve as a critical indicator Table 3 presents an overview of the indicators and a detailed description of nursing students' self‐directed Learning Effectiveness, which are presented by relevant subtheme.

Table 3.

Dimensions of perceived self‐directed learning outcomes through AI‐based learning.

Sub‐themes	Perceived dimensions	Description of self‐reported experiences
Information Literacy and Self‐Management Skills	Information identification and evaluation skills	Students reported employing strategies to identify and evaluate the authority and reliability of learning resources, perceiving an improved ability to analyse source credibility
	Resource acquisition and retrieval efficiency	Students described utilising AI to quickly locate materials and optimise search strategies, which they felt minimised time wastage
Critical and innovative thinking skills	Critical thinking ability	Students reported engaging in independent critical thinking by assessing the logical coherence and reasonableness of AI‐generated content
	Innovative thinking ability	Students noted that AI served as a catalyst for innovative thinking, inspiring perception, logical reasoning, and creative exploration of new solutions.
Perceived readiness for clinical decision‐making	AI‐assisted procedural preparation	Students reported utilising AI tools to review procedural steps, which they perceived as enhancing their confidence and mental preparation for clinical adaptability
	Case‐based reasoning exercises	Students described using AI tools to simulate decision‐making exercises, feeling that it enhanced their clinical judgement.

3.5.1. Subtheme 1: Information Literacy and Self‐Management Ability

In the context of SRL, AI acts as both an enabler and a challenge for information processing. Faced with voluminous AI outputs, students are forced to elevate their information literacy.

In fact, the information given by AI is very long and comprehensive, so I must extract from it. That is, the content related to the disease I am dealing with, the teacher's question, and the project I am working on. I cannot say it answers the parts I don't want or the parts I already know. I must filter out the parts I don't know.

(F8)

When evaluating information, students must locate and retrieve useful learning content efficiently and accurately. This task tests their time and resource management capabilities, or overall self‐management skills.

Retrieve relevant information on clinical issues through keyword matching to ensure content accuracy.

(F6)

Generally, AI‐provided information includes main titles and subheadings. When I filter information, I usually first scan through the directions and titles it lists, and then focus on the content relevant to me.

(F10)

So, within the learning objectives it sets, if I identify certain parts as my key focus areas or particularly challenging topics, I will allocate more time to study them…if the AI assigns two hours2 h to study a difficult point A and another two hours2 h for difficult point B, during the learning process, I might find that difficult point B is actually not as challenging for me.

(F3)

3.5.2. Subtheme 2: Critical and Innovative Thinking Ability

Some interviewees noted that critical thinking underpins AI‐based learning. Through regular reflection and self‐assessment, they strive to avoid excessive dependence on such tools and attach great importance to independent thinking and self‐development.

When I have my own thoughts, I won't deliberately use AI. I only rely on it when my thoughts are confused. I don't want it to replace human spontaneous thinking

(F1)

AI is highly effective as a tool for information extraction and integration, saving time on initial preparations and allowing you to focus on tasks that require your own thinking. However, it's advisable not to rely on it too heavily. You should have the ability to critically assess the feedback it provides. Given the current limitations of AI development, it cannot completely replace your work. It's important to identify your unique, irreplaceable value and enhance your personal worth.

(F5)

Other participants observed that when they encounter a mental block, AI functions as an exceptionally valuable tool, offering them boundless inspiration and subtly augmenting their innovative thinking abilities.

The AI image generation function is quite interesting. For instance, if I read a book and ask it to generate pictures based on the scenery in the book, it can help me better understand the text and bring me new ideas.

(F11)

For example, when using Meta‐analysis, in the topic selection stage, it will preliminarily screen according to specific requirements and relevant journal content. At that time, I had no clear direction, so I could hand over this problem to AI to see which aspects were more innovative.

(F19)

3.5.3. Subtheme 3: Perceived Readiness for Clinical Decision‐Making

Clinical settings feature complex, changing situations for nursing students. Participants reported that AI tools enhanced their comprehension of clinical cases and better prepared them for practical nursing work.

This also fosters a sense of proactive engagement. For instance, regarding the proficiency of certain procedures, AI can provide feedback and correct me on aspects where I may not have been thorough or precise. It doesn't merely follow standardised approaches…For my upcoming clinical practice, this actually provides me with a highly structured and standardised template. I can familiarise myself with it in advance, understand key points to pay attention to, and feel more confident in my future clinical practice

(F23)

3.6. Theme 4: Adaptive Challenges

Indeed, several interviewees recounted experiencing challenges stemming from their engagement with AI‐based learning, which posed difficulties for them.

3.6.1. Subtheme 1: Delayed Iterative Updating of AI Knowledge Bases

AI guidelines cannot keep pace with dynamically changing clinical contexts due to update lags. Worries about obsolete knowledge affect nursing students' clinical decision‐making and self‐directed learning motivation.

For example, with IV infusion, AI might offer a comprehensive, step‐by‐step process when asked how to perform it. Yet, in real‐world clinical practice, further assessment is necessary based on the patient's specific condition, whether they are obese, underweight, edematous, elderly with poor veins, infants, or post‐surgical patients with significant bleeding. AI may not fully grasp these complex, real‐life clinical situations. Therefore, I believe it is essential to apply the specific, context‐appropriate procedures used in clinical practice, tailored to each individual case.

(F21)

In a clinical setting, practices might be considered more advanced, or perhaps they might seem slightly outdated compared to what AI offers. The reasons behind this could vary and might require further examination. For instance, even if a guideline is newly released, it might lack substantial data to support its recommendations.

(F16)

3.6.2. Subtheme 2: Issues Related to Efficiency and Learning Depth

AI provides immediate feedback, leading many nursing students to prioritise efficiency and adopt fragmented learning. Most focus on superficial content rather than in‐depth exploration. This undermines their systematic knowledge integration capacity and creates a conflict between learning efficiency and academic depth.

It might make me lazy and inclined to cut corners, as if having AI means I can skip any important steps. However, in most cases, these steps require reflection and reevaluation. You need to think critically about the feedback AI provides and refine it further to truly grasp the overall understanding of the task

(F5)

4. Discussion

This study offers a comprehensive understanding of nursing students' AI‐based learning experiences. Anchored in Bandura's SCT [15, 16] and Zimmerman's SRL theory [13, 14], it clearly articulates how these findings support, extend, and challenge existing theoretical assumptions. Our findings support the SCT construct of self‐efficacy and the forethought phase of SRL. Specifically, within the SRL framework, while AI significantly optimises this forethought phase by helping students plan and set goals efficiently, it also interferes with the crucial self‐reflection phase. However, AI tools simultaneously pose risks by fostering emotional dependence, bypassing critical self‐reflection and encouraging cognitive reliance rather than in‐depth knowledge processing. Accordingly, this study extends the application of SCT in digital environments by revealing a complex outcome: AI can raise individuals' perceived self‐efficacy in the short run, but erodes their independent cognitive resilience over time. Experiences of nursing students demonstrate that while boosting autonomous learning efficiency, AI‐based learning carries hidden learning risks.

To properly contextualise these behavioural shifts, it is essential to acknowledge the unique institutional landscape of Chinese nursing education. During their final year, nursing students in China undergo a rigorous clinical internship while simultaneously preparing for highly competitive graduate entrance or national licensing examinations [21]. This intense dual pressure creates a high‐stress educational ecology characterised by severe time constraints. Consequently, an urgent demand for extreme learning efficiency emerges, driving students to rely heavily on AI tools to rapidly process voluminous academic and clinical information. Understanding this specific context naturally explains why fragmented knowledge acquisition surfaces as a primary adaptive challenge. Driven by the need for immediate academic survival, students are often compelled to prioritise rapid answers and cognitive shortcuts over the deep, systematic comprehension necessary for complex medical training [22].

When compared to existing evidence, our findings both extend and diverge from the current body of literature. Recent global discourse has recognised the educational potential of generative AI like ChatGPT, which helps nursing students acquire medical knowledge, obtain tailored learning guidance and boost overall learning efficiency [23]. In terms of emotional cognition, prior studies consistently report favourable dispositions and increased confidence toward digital learning tools among nursing students [24, 25]. Built on SCT, this study further supplements existing evidence and reveals a distinct vulnerability behind this heightened self‐efficacy. The hyper usability provided by AI tools removes the productive struggle traditionally required in profound learning. Consequently, students develop emotional dependence, expressing fears about a potential decline in independent critical thinking if detached from the technology. Furthermore, whereas existing international literature predominantly praises generative AI for providing personalised learning paths [8, 26], our findings diverge based on actual clinical exposure. Participants noted that AI‐generated standardised guidelines often fail to account for highly dynamic real‐world clinical variables, such as intravenous infusion for elderly patients with poor veins and other special patient groups. This underscores that while AI serves as an efficient theoretical tutor, its current iteration suffers from delayed knowledge updating and lacks the contextual nuance required for advanced clinical reasoning [27]. The observed fragmented learning is consistent with findings from existing scholarly research [28]. Students fail to develop systematic clinical reasoning and instead adopt a rigid checklist mentality that lacks sufficient clinical depth. In dynamic healthcare contexts, such fragmented knowledge acquisition hinders the development of independent problem‐solving competencies essential to professional nursing practice.

Regarding clinical decision‐making, it is necessary to distinguish between subjective perceptions and objective performance. While students reported feeling more prepared and confident in reviewing procedural skills, these represent self‐reported cognitive shifts rather than objective measures of clinical competence. Psychological preparedness is undoubtedly a crucial precursor to clinical readiness [29]. Yet, actual clinical proficiency still necessitates hands‐on psychomotor practice, authentic patient communication, and validation through objective clinical assessments.

Drawing on these findings, specific implications can be proposed for educational practice and technological development, as well as the refinement of AI‐supported instructional strategies in nursing education. To meaningfully integrate AI into teaching and learning processes, actions must be taken at both the institutional and pedagogical levels. First, at the institutional level, nursing colleges must establish clear, ethical guidelines regarding acceptable AI usage, creating a psychologically safe environment where students can use AI for brainstorming without fear of academic penalty. Second, at the pedagogical level, educators should shift from passive acceptance of technological learning to delivering targeted, skill‐oriented learning guidance [30]. Curricula should integrate error‐analysis tasks that require students to apply critical thinking to identify intentional clinical flaws within AI‐generated care plans [31]. This strategy directly addresses emotional dependence by forcing the crucial self‐reflection phase of SRL that automation frequently bypasses [32]. Simultaneously, design teams need to recognise that generic language models are insufficient for advanced medical education. Developers ought to prioritise specialised AI platforms for nursing with transparent reference tracing functions [33]. To address the outdated content update issue highlighted in this study, such tailored educational systems must operate dynamically and maintain timely content iteration.

5. Limitations

This study has certain limitations that should be acknowledged. First, single‐institution sampling: participants were purposively recruited from a single nursing college at a medical university in China. The similarity in their educational backgrounds, heavily influenced by the local intensive internship model, may limit the transferability of the findings to different cultural or institutional contexts. Second, potential self‐selection bias: students who volunteered to participate in this study may possess a higher baseline interest or greater proficiency in using AI technologies compared to their peers. This could potentially skew the findings toward a more favourable or advanced usage perspective than that of the general nursing student population. Finally, the rapidly evolving nature of AI technologies (e.g., the frequent updates of platforms like ChatGPT, Doubao, and DeepSeek) means that this study captures a specific temporal snapshot of AI experiences. As these tools become more sophisticated, the generalisability and longevity of our specific adaptive challenges (such as delayed knowledge updating) may shift. Future research should expand geographically, employ quantitative designs to measure the objective impact of AI, and conduct longitudinal tracking of AI integration.

6. Conclusion

This study illustrates the dual impacts of AI‐based learning among nursing students. Viewed through the lenses of SRL and SCT, AI tools profoundly enhance perceived clinical readiness and learning efficiency, yet simultaneously introduce adaptive challenges such as fragmented learning and emotional dependence. By distinguishing perceived proficiency from objective clinical competence, this research provides targeted, practical implications. For nursing students to safely navigate the evolving healthcare landscape, educational strategies must proactively cultivate critical evaluation skills, and developers must tailor AI tools to the rigorous, evidence‐based demands of clinical nursing.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

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