Abstract

Background:

Digital affordances refer to the possibilities provided by digital environments for learners. In the context of nursing education, artificial intelligence (AI) chatbots currently offer multimodal learning approaches and demonstrate various possibilities for digital actions. Therefore, exploring the digital affordances of AI chatbots in nursing education is crucial for the continuous advancement of the field.

Objective:

To evaluate the digital affordances of AI chatbots in nursing education, focusing on the relationship between digital affordances and learning gains.

Methods:

We employed affordance theory to conceptualize the potential actions of AI chatbots and utilized a taxonomy of affective, behavioral and cognitive learning gains to conduct a systematic review in nursing education.

Results and conclusions:

A total of 25 studies were identified in this systematic review. The geographical distribution of the studies is mainly in Asia. The most used study designs were quantitative designs (n = 12) with sample sizes between 16 and 457. The duration of these studies is usually short, ranging from a few hours to 3 months. The included studies reported several digital affordances of AI chatbots in nursing education, including assistance provision, personalization, human-like conversing, distilling information, and fostering familiarity. However, four digital affordances-facilitation, enriching information, context identification, and ensuring privacy-still lack empirical support. The evidence for the digital affordances of AI chatbots in nursing education was dominated by cognitive learning gains (such as learning achievement, critical thinking, and problem solving) and followed by affective (such as learning interest, self-efficacy, and enjoyment) and behavioral learning gains (such as engagement, diagnostic skills and clinical practice). However, several studies reported no statistically significant improvement in certain cognitive learning gains, particularly knowledge acquisition and clinical reasoning competency. Similarly, limited evidence was found for improvements in learners’ confidence and satisfaction. These findings suggest that the current evidence remains inconclusive. Future research should employ longer study durations and larger sample sizes to further examine the educational impact of AI chatbots.

1 Introduction

With rapid advances in generative pretraining transformer (GPT) technology, the application of AI chatbots in nursing education has opened up new possibilities for goal-oriented actions (1–4), sparking extensive academic discussions on their innovative applications (5–7). Previous studies have revealed that AI chatbots facilitate well-designed multimodal conversations that can effectively assist in a variety of areas such as patient intervention (8), postoperative support (9), and psychiatric care (10), which has attracted great interest in nursing education research (3). On the other hand, concerns have been raised about the use of chatbots in nursing education. These concerns include: since ChatGPT responses are based on patterns learned from data, issues of accuracy, reliability, and potential bias remain a concern (11); there has been a lack of content and programs for instruction using chatbots (12); and there are concerns about over-reliance on technology, such as issues with plagiarism and limiting critical thinking skills (13). The application of AI chatbots has significantly expanded the scope of nursing education, offering students new goal-oriented action possibilities (i.e., affordances) (14). However, there remains a gap in research regarding the identification of these affordances and their practical applications in nursing education. In other words, how to effectively identify digital affordances in nursing education and in what nursing educational contexts these affordances will emerge remain unclear.

Previous literature has indicated that AI chatbots in nursing education provide nursing students with various digital learning opportunities (15). For example, Han et al. (16) reported that using AI chatbot programs in electronic fetal monitoring significantly improved nursing students’ self-directed learning and nursing skills. Furthermore, Shorey et al. (17) found that simulating patient interaction scenarios with AI chatbots helped nursing students strengthen their communication skills. Another study demonstrated that AI chatbots enhanced students’ learning performance and self-efficacy in obstetric vaccine education by engaging with case-based scenarios (18). In other words, AI chatbots not only support the personalized learning but also promote students’ learning behaviors and skill development, creating new learning opportunities and facilitating the conversion of these opportunities into learning gains.

Despite these potential benefits, the digital affordances of AI chatbots in nursing education still face several challenges. First, students’ cognitive abilities, technology acceptance, and trust in AI chatbots affect the realization of their affordances, as students may not fully understand the opportunities and technical support offered by AI chatbots (12). Second, integrating AI chatbots into traditional nursing education is involved complex and dynamic interactions, and the interplay between teachers, students, and AI chatbots remains insufficiently clarified (19, 20). Lastly, it remains uncertain to what extent the digital affordances of AI chatbot contributes to measurable learning gains, and there is a need to analyze and distinguish these learning gains and their assessment methods (21, 22). Therefore, it is crucial for exploring the new goal-oriented possibilities created by AI chatbots for nursing students and how the digital affordances affect the learning gains.

Due to the increasing popularity of AI chatbots, some empirical studies have explored their effects on students’ learning gains (4). However, there is a lack of in-depth analysis of the similarities and differences in learning gains across different contexts, which raises new questions regarding how to meaningfully and practically assess cognitive, affective, and behavioral learning gains (22–24). For example, Chang et al. (25) demonstrated that AI chatbots not only enhance the cognitive learning performance of nursing students but also improve their behavioral self-efficacy. Additionally, with the rapid advancements in nursing research and the advent of new diseases and medications, nursing professionals are challenged to quickly adapt to evolving medical knowledge, make accurate decisions in real-case scenarios, and master complex nursing skills (1, 17, 26–28). Therefore, incorporating valid and reliable measures of learning gains in nursing education is crucial to ensuring that nursing students effectively master fundamental knowledge, develop critical thinking, enhance problem-solving abilities, and improve communication skills (16, 25, 26). Several studies have utilized AI chatbots in various methods to assess the quality of nursing education (25), such as Han et al. (16), who employed AI chatbots for assessing skills in electronic fetal monitoring, and Shorey et al. (17), who used AI chatbots in patient communication simulations. While these approaches are promising, there remains a need for more robust empirical evidence to validate their effectiveness in accurately measuring the diverse learning gains of nursing students.

Compared with prior reviews on AI chatbots in education, this study offers several distinct contributions. To better understand and interpret the complex interactions between nursing professionals and AI chatbots, this study adopts the perspective of affordance theory (29, 30), conceptualizing the potential actions that AI chatbots enable for educators and students. Concurrently, we have employed a coherent classification approach proposed by Rogaten et al. (24) to review various learning gains and analyze their similarities and differences. In addition, AI literacy is incorporated to account for nursing students’ knowledge, skills, and attitudes required to effectively understand, evaluate, and interact with AI chatbots (31). This approach aims to develop a more comprehensive understanding of the learning gains reported in the literature and to assess the potential of these learning gains as measure of the value of nursing education. The study aims to answer the following research questions:

RQ1: What digital affordances do AI chatbots provide in nursing education?

RQ2: What types of evidence have been employed to assess the impact of AI chatbots on learning gains in nursing education?

2 Method2.1 Search strategies

The most recent search was conducted in the period of November–December 2025 in PubMed and Web of Science (see Table 1). Keywords included combinations of “chatbot,” “nursing,” and “education.” We developed our searching teams based on previous studies, which suggested that nursing education programs may include both health professionals and nurses. Moreover, chatbots in educational programs have been described using various terms, including “Conversational Agents” (32), “AI agent” (33), “artificial intelligent agent” (34) and “ChatGPT” (35). Therefore, we expanded our search in PubMed and Web of Science to include broader terms: “chatbot” OR “Conversational Agent” OR “AI agent” OR “artificial intelligence agent” OR “ChatGPT” (Topic) AND “nurse” OR “nursing” OR “care” OR “health” (Topic) AND “education*” OR “train*” (Topic). We retrieved 178 studies from PubMed and 143 studies from Web of Science.

DatabaseSearch termsRetrievedPubMed(“chatbot” OR “Conversational Agent” OR “AI agent” OR “artificial intelligence agent” OR “ChatGPT”) and (“nurse*” OR “nursing*” OR “care” OR “health”) and (“education*” OR “train*”)179Web of Sciencechatbot or Conversational Agent or AI agent or artificial intelligence agent or ChatGPT (Topic) AND nurse or nursing or care or health (Topic) AND education* or train* (Topic)144

Search queries and number of results for each database.

2.2 Inclusion criteria

The inclusion criteria for this study were established using the PIOS framework (see Table 2), which is adapted from the widely used PICOS (participants, interventions, comparators, outcomes, and study design) strategy (36). This decision is justified by the nature of the included studies, many of which do not involve a clearly defined comparator group. Participants include nurses and nursing students. Interventions involve the use of AI chatbots or similar dialogue systems in nursing education, focusing on their impact and effectiveness. The outcomes are concerned with the utilization of AI chatbots as educational or training tools and their effectiveness in nursing education. Study designs encompass all research related to nursing education or training. No direct comparator group was specified.

Inclusion criteriaDescriptionParticipantsNurses, nursing studentsInterventionsInvolving AI chatbots or similar dialogue systemsOutcome(s)Utilization of AI chatbots as an educational or training toolStudy designsInvolving nursing education and in service training or professional development2.3 Search process

We conducted a rigorous screening and selection process for this study. Initially, our search identified 323 studies. After removing 88 duplicates, we had 235 unique records. During the title and abstract screening phase, we excluded 205 records for two primary reasons: 188 did not target our demographic of nurses and nursing students, and 17 were review articles. Subsequently, we thoroughly reviewed the full texts of the remaining 30 studies for eligibility. Of these, 5 studies were excluded as they did not focus on nursing education or training. Ultimately, 25 studies met our comprehensive inclusion criteria and were selected for inclusion in the study (see Supplementary file 1).

Detailed methodological characteristics of the included studies, including sample size, intervention duration, and research design, are reported in Supplementary file 1. Further information on the study selection process is provided in the PRISMA flow diagram (Figure 1).

PRISMA flow diagram for systematic review (67).

2.4 Data analysis and synthesis

The coding system in this study was inspired by Mygland et al. (29) review of state-of-the-art literature on human-AI interactions, which identified and categorized 91 different digital affordances of AI chatbots into nine high-level classes. To improve transparency, a simplified version of the coding framework is presented in Table 3 for reviewers. This table summarizes the main affordance categories used in the analysis. More detailed definitions and coding rules are provided in Supplementary file 2. Furthermore, we applied the ABC learning gains classification approach to analyze digital affordances of AI chatbots across three dimensions: (1) affective learning gains, which can be defined as a change in affect related states during a course (24), illustrated by Han et al. (16) in their evaluation of an AI chatbot’s affective impact on nursing students and nursing trainees’ skills through in a quasi-experiment; (2) behavioral learning gains, which focus more strongly on skills than knowledge (24), demonstrated by Hsu (11) in their study on enhancing medical terminology learning using ChatGPT and Termbot; and (3) Cognitive learning gains, which can be defined as development in knowledge, understanding and cognitive abilities (24), as evidenced by Chang et al. (25) use of a chatbot in a quasi-experiment to improve obstetric vaccination learning, cognitive learning gains.

High-level affordancesDefinitionReferencesHuman-like conversingChatbots use AI to simulate conversation, and this changes how nursing students and trainees use software. These tools can understand user intent, make responses, and remember the chat history for later questionsLunberry and Liebenau (68), Stoeckli et al. (14), Waizenegger et al. (69), Moussawi (70), Lippert et al. (71)Assistance provisionChatbots assist in daily tasks and nursing education, offering reminders, functions, info, admin, and personalized support to boost engagement and efficiencyStoeckli et al. (14), Stoeckli et al. (72), Waizenegger et al. (69), Knote et al. (73), Moussawi (70), Meske and Amojo (74), and Barnett et al. (75)FacilitationChatbots connect nursing students and trainees with organizations, unify system access, and simplify tasksStoeckli et al. (14), Knote et al. (73), and Meske and Amojo (74)Distilling informationChatbots help summarize information and support mood reflectionStoeckli et al. (14), Stoeckli et al. (72), Knote et al. (73), and Meske and Amojo (74)Enriching informationChatbots use visuals and extra text to share information faster and build stronger connectionsStoeckli et al. (14) and Knote et al. (73)Context identificationChatbots understand the context of a conversation, identify needs, give helpful responses, and guide the discussionStoeckli et al. (14), Stoeckli et al. (72), Knote et al. (73), Meske and Amojo (74)PersonalizationChatbots adapt to nursing students and trainees, personalizing responses and toneWaizenegger et al. (69), Knote et al. (73), Moussawi (70), and Lippert et al. (71)Fostering familiarityChatbots in nursing education build on users’ familiarity with chat apps. Students feel at ease with this interaction, yet unmet expectations may cause dissatisfactionMoussawi (70)Ensuring privacyChatbots protect privacy and control access. Users share information, so interactions require careful managementStoeckli et al. (14), Stoeckli et al. (72), Waizenegger et al. (69), and Knote et al. (73)

High-level digital affordances identified in the literature (29).

3 Results

This section addressed the aims of the present study in the following order: (1) characteristics of study (including demographic, AI chatbot technology, and research design) (see Table 4); (2) investigated the digital affordances provided by AI chatbots in nursing education; and (3) empirical evidence obtained through the evaluation of AI chatbots.

N = 25NSample sizeLess than 20120 to 59560 or more participants10No information available9Intended userJunior nursing students7Senior nursing students14Nurse practitioners1No information available3Country/RegionUSA6Taiwan5Spain3Singapore2UK1Mainland China1Türkiye1Canada1Hungary1South Korea1Italy1Romania1Germany1AI chatbot technologyLarge language model10NLP9Voice-controlled AI assistants2Knowledge-based system1Pattern matching1No information available2AI chatbot durationLong term (monthly)6Short term (daily)5Medium term (weekly)4No information available10Research designQuantitative (quasi-experiment, cross-sectional comparative study, descriptive statistics, questionnaire)12Qualitative (interviews)2Mixed methods (document-analysis, usability tests, cross-sectional technology validation study)4No information available7

Characteristics of study.

3.1 Demographic

Overall, the sample size of these empirical studies tends to be small, ranging between 16 and 457. The sample sizes in the study were categorized into four groups: less than 20 participants (n = 1), 20 to 59 participants (n = 5), 60 or more participants (n = 10), and a category (n = 9) where specific information was not provided. The intended user groups for the AI chatbots included nursing students and nursing professionals. Of the 25 studies, 7 focused on junior nursing students, 14 on senior nursing students, and 1 on nurse practitioners. However, 3 of these studies did not specify their intended user groups. The geographical distribution of the studies was global, with a focus in Asia (5 from Taiwan, 2 from Singapore,1 from Mainland China, 1 from South Korea), North America (6 from the USA, 1 from Canada), and Europe (3 from Spain, 1 from the UK, 1 from Türkiye,1 from Hungary, 1 from Italy, 1 from Romania, 1 from Germany). This international representation highlights the global interest in AI chatbot technology and its applications across different healthcare systems and educational backgrounds. The research encompasses a broad range of user groups, from junior nursing students to senior nursing students and nurse practitioners, demonstrating the potential for AI chatbot applications at various levels of nursing education.

3.2 AI chatbot technologies and duration

The results indicate that the usage scenarios of AI chatbots involve dialogue facilitation, intent understanding, and providing instant feedback (37), with both short-term and long-term interventions potentially improving learning gains. Consequently, the AI chatbots in the studies were primarily powered by technologies such as large language model (n = 10), NLP (n = 9), voice-controlled AI assistants (n = 2), knowledge-based systems (n = 1), and pattern matching (n = 1). Two studies did not clearly specify the key technology used. The duration of these studies varied, ranging from 3 months (17) to a few hours (16). The implementation of AI chatbot applications was categorized into long-term, medium-term, and short-term: 6 studies used the AI chatbot monthly (long-term), 5 studies daily (short-term), and 4 studies weekly (medium-term). Ten studies did not provide specific information regarding the duration of the experiment.

3.3 Research design

In reviewing 25 studies on the application of AI chatbots in nursing education, we found that quantitative studies (n = 12) primarily addressed the effectiveness of AI chatbots in enhancing nursing skills and knowledge, including medical terminology learning, physical examinations, and electronic fetal monitoring (11, 16, 25). Most of these studies (n = 7) employed a quasi-experimental design to assess learning gains using validated scales. In contrast, two qualitative studies measured learning gains through focus group interviews, exploring nursing students’ perceptions, attitudes, and experiences with AI chatbots (17, 26). Mixed-methods studies (n = 4) provided an opportunity to evaluate affective, behavioral, and cognitive learning gains, such as in studies examining the impact of AI chatbots on students’ self-regulated learning abilities (28).

3.4 The digital affordances of AI chatbots for students and educators in nursing education

This review of 25 studies reveals five high-level digital affordances provided by AI chatbots in nursing education: assistance provision, personalization, human-like conversing, distilling information, and fostering familiarity, as detailed in Table 5. We conceptualize digital affordance as “the possibility of goal-oriented action offered to a specific group of users through a technological object” (38). The concept of digital affordance is therefore relevant and takes into account (1) the competencies and learning goals of nursing students and (2) the characteristics of the AI chatbot (39).

High-level digital affordancesDefinition and identified affordancesReferencesAssistance provisionAI chatbots can perform a variety of assistive tasks, including: (1) speedy assistance, (2) usefulness, (3) executing tasks, (4) live updates, (5) data access, (6) quick answers, (7) error reduction, (8) support seekingHan et al. (16), Otero-Agra et al. (76), Chang et al. (18), Chang et al. (25), Chang et al. (19), Chow et al. (27), Hsu (11), Chen et al. (26), Sáiz-Manzanares et al. (28), and Kaur et al. (77)PersonalizationAI chatbots can adapt interactions to their users by providing customized responses and adjusting their tone and style, including: (1) personalized learning, (2) interactivity, (3) feedback, (4) adaptivity.Chang et al. (18), Hsu (11), Hsu et al. (78), Chen et al. (26), Riedel et al. (44), Chang et al. (19), Han et al. (16), and Saiz-Manzanares et al. (28)Human-like conversingAI chatbots can generate human-like messages, including: (1) human-like content, (2) conversation mimicryChow et al. (27) and Chen et al. (26)Distilling informationAI chatbots provide users with action possibilities related to distilling information, including: (1) flow maintenance, (2) aggregated dataHan et al. (16), Chang et al. (25), Kaur et al. (77), and Chang et al. (19)Fostering familiarityAI chatbots allow user to express their needs directly through a familiar interaction mode, including: (1) emotional connection, (2) comfort growthChow et al. (27), Chen et al. (26), and Kaur et al. (77)

High-level affordances provided by AI chatbots related to nursing education.

In this context, the most common high-level digital affordance provided by AI chatbots is assistance provision, with eight related affordances identified (see Supplementary file 3). Conversely, human-like conversing, distilling information and fostering familiarity were less frequently mentioned. However, the role of AI chatbots seems to focus more on individual support rather than facilitating traditional human-to-human interactions in educational settings. Therefore, the use of AI chatbots in nursing education appears to prioritize practical needs such as assistance provision and personalization over simulating human-like conversing, distilling information and fostering familiarity. It may be worthwhile to consider integrating AI-based chatbots into nursing learning environments to enhance students’ metacognitive awareness and the quality of peer feedback. For instance, researchers have reported that AI-based chatbots can significantly improve learners’ metacognitive awareness (40). In one study, an AI chatbot named Eva was integrated into an online peer review system, significantly improving the quality of student feedback (41). This indicates that AI chatbots may not be fully utilized to offer more interactive and engaging educational experiences.

3.5 Relationships between digital affordances and learning gains

An analysis of 18 related affordances revealed an uneven distribution of evidence across learning gains (see Table 6). Specifically, 4 of the 18 related affordances (22.20%) assess a combination of all three learning dimensions (affective, behavioral, and cognitive) (see Figure 2). In contrast, 5 related affordances (27.80%) focus on both affective and cognitive learning gains, 2 related affordances (11.10%) on affective and behavioral, 2 related affordances (11.10%) on behavioral and cognitive, 1 related affordance (5.60%) exclusively on affective learning gains, 1 related affordance (5.60%) on behavioral learning gains, and 3 related affordances (16.60%) on cognitive learning gains. However, there are fewer learning gains have been studied for behavioral change, but some researchers believe that behavioral learning gains deserve further study and that skill development and engagement are as important as knowledge acquisition (37, 42, 43).

High-level digital affordancesRelated affordances identified in nursing educationAffectiveBehavioralCognitiveAssistance provisionUsefulness✓✓✓Speedy assistance✓✓Executing tasks✓Live updates✓✓✓Data access✓✓Quick answers✓Error reduction✓Support seeking✓✓PersonalizationPersonalized learning✓✓✓Interactivity✓✓✓Feedback✓✓