Exploring nursing students’ communication experiences with artificial intelligence–supported virtual patients: a qualitative study

Abstract

Background

Effective communication is essential in nursing, and artificial intelligence (AI) tools such as ChatGPT are increasingly integrated into nursing education.

Methods

This phenomenological study explores how ChatGPT-supported virtual patient scenarios influence nursing students’ empathy and communication skills. Third-year nursing students enrolled in a Family Health Nursing elective interacted with a virtual patient through predefined ChatGPT prompts simulating a counseling session. Data were collected through focus group interviews and online reflective diaries.

Results

Thematic analysis revealed three themes: “Unrealistic empathy and communication experience”, “Solution-oriented and effective communication experience”, “Communication and empathy awareness: Future communication strategies and professional preparation”

Conclusions

Although participants perceived ChatGPT-supported virtual patient communication as providing a less realistic communication and empathy experience than real patient interactions, they also reported that ChatGPT’s interaction style offered a structured and solution-oriented communication experience. Furthermore, the ChatGPT-supported virtual patient communication experience was perceived as helping students make reflection and self-assessment about communication and empathy, and identify areas for improvement before interacting with real patients. These findings suggest that ChatGPT-supported virtual patient communication cannot replace real patient encounters, but it can serve as a complementary educational tool that encourages reflective learning and prepares nursing students for subsequent clinical communication experiences.

Introduction

Communication is a process that facilitates mutual understanding by exchanging feelings, thoughts, and information between individuals [1]. Nurses must have practical communication skills because nursing is a profession where individuals constantly communicate and provide care. Communication includes skills such as verbal and non-verbal expressions, active listening, and empathizing [2]. Empathy helps nurses understand patients’ situations better and makes patients feel valued. Empathic communication is particularly critical in sensitive contexts such as end-of-life care, where it directly affects families’ experience [3].

Empathy and communication skills are recognized as teachable, and structured educational programs can effectively develop them [4]. Improved communication skills among nursing students have been associated with a greater ability to deliver patient-centered care [5]. Technology-based education programs offer an innovative approach in teaching communication skills. Studies have demonstrated that digital tools and virtual patient scenarios improve nursing students’ communication and empathy skills. In virtual simulations via Zoom, nursing students practiced empathy and communication by informing family members of a rapidly deteriorating patient’s condition, and this experience was associated with improvements in students’ clinical decision-making skills [6]. Simulation-based interventions have been associated with improvements in nursing students’ empathy skills. A study conducted in South Korea reported that simulation-based training was associated with increased cultural empathy levels among students [7]. Systematic reviews suggest that such interventions can improve student nurses’ communication and empathy levels [8, 9].

In recent years, the use of artificial intelligence-based applications, particularly tools like ChatGPT (OpenAI, San Francisco, CA, USA), has increased significantly in nursing education. According to a study, ChatGPT-supported virtual patient simulations have been shown to improve nursing students’ empathy and communication skills [10]. In a separate study, AI-powered virtual health communication (AI-VHCom) was associated with improvements in students’ emotional, cognitive, and communication skills [11]. Similarly, Kim et al. [12] stated that GPT-based virtual patient simulations support students’ health assessment and communication skills, and that students particularly value the possibility of unlimited repetition as a significant advantage. These simulations provide students with an easily accessible, cost-effective and interactive learning environment. A recent review synthesizing forty studies on large language model-based virtual patient simulations in medical and nursing education identified emotional realism, alongside factual accuracy and role consistency, as a persistent challenge across current implementations [13]. However, the literature on ChatGPT-supported virtual patient applications remains recent and limited, and has mainly focused on students’ experiences, acceptability, and basic communication skills. Consequently, little is known about how nursing students subjectively experience and construct empathy within an unstructured, text-based AI interaction, particularly within the context of a structured undergraduate course. To address this gap, the present phenomenological study explores nursing students’ empathy and communication experiences with ChatGPT-supported virtual patient scenarios within the Family Health Nursing course, aiming to clarify the potential role of such digital tools in nursing education from the students’ own perspective. This study focuses on students’ subjective, perceived experience of empathy and communication during the interaction, rather than assessing their actual implementation of empathetic communication skills or the pedagogical effectiveness of the intervention.

Methods

This research is a qualitative phenomenological study. Phenomenology aims to understand how participants perceive an experienced phenomenon [14].

Participants and setting

Purposive sampling was used to select participants from third-year nursing students enrolled in the Family Health Nursing course at a university in Turkey’s Mediterranean region.

Ethical considerations

Ethical approval was granted by the Akdeniz University Clinical Research Ethics Committee (TBAEK-740). Participants were informed that the recordings would be used solely for scientific purposes and provided written consent. All participants were identified by the code “NS” (Nursing Students). The study was conducted in accordance with the ethical principles of the Declaration of Helsinki.

Implementation

The virtual patient interaction was conducted on November 20, November 27, December 4, and December 11, 2024. These dates coincided with the weeks during which students delivered their group presentations within the scope of Family Health Nursing course and were therefore selected for implementation. Each session was planned to include four student participants; however, in one session, only three students volunteered. One participant was unable to participate in the application process due to fatigue.

ChatGPT, a generative artificial intelligence (GenAI) tool developed by OpenAI, was used to facilitate the virtual patient interactions. The virtual patients scenarios were developed by the Family Health Nursing course instructor and the second researcher using ChatGPT. The scenario topics included rational drug use, family health in dual-income households, and health management during the pandemic. Biographical data, medical history, and case details were entered into ChatGPT to generate draft scenarios, which were then reviewed for alignment with the course content, appropriateness for students, and clarity. The scenarios required no additional editing.

Prior to the implementation, ChatGPT applications, where students could interact as a virtual patient, were opened through two different account sessions for the researchers. Then, scenarios were uploaded to ChatGPT. ChatGPT was instructed to assume the role of a patient seeking consultation from a Family Health Nurse according to scenario (Fig. 1).

Fig. 1.

ChatGPT-supported virtual patient communication procedure

In the implementation, participants adopted the role of Family Health Nurses. These sessions are described as virtual patient interaction. Each student, as a Family Health Nurse, interacted with their virtual patient on ChatGPT in a separate, 20- to 25-minute. The activity was conducted in the Communication Laboratory, where participants engaged in one-on-one interactions with a virtual patient via ChatGPT.

Consistency across simulations was promoted by conducting all virtual patient interactions under standardized conditions, including the same environment, individual computers for each participant, no prior access to the scenarios, and the use of the same prompt before each ChatGPT session.

After the session, a focus group interview was held, and participants completed an online journal within 24 h (Fig. 1).

Data collection

Data collection occurred after the program concluded. The diaries included guiding questions and prompts that facilitated deeper participant reflection and detailed descriptions of emotions and learning experiences (Table 1). Additionally, the diaries captured experiences not discussed in focus groups, thereby increasing the variety and depth of the data.

Table 1.

Reflective journal guide questions and focus group interview questions

Reflective journal guide questions: 1. What did today’s experience in ChatGPT-assisted virtual patient scenarios teach you?  ● Describe your feelings about this experience. What emotions did it evoke in you? 2. What challenges did you encounter in communicating with the patient, and how did you overcome them?  ● Why was this event significant for you?  ● Describe your feelings about this experience. What emotions did it evoke in you? 3. In retrospect, is there anything you could have done differently in the scenario? 4. How has this experience influenced your professional or personal development? 5. How will this experience shape your future interactions with patients? 6. How has this experience challenged or altered your values, beliefs, or assumptions regarding patient communication? 7. How will this experience impact your values, beliefs, or assumptions regarding empathy toward patients? 8. How will this scenario influence your ability to practice empathy in real-life situations?

Focus group interview questions:1. Can you describe your experience with communication in ChatGPT-assisted virtual patient scenarios? 2. Can you describe your empathy experience in ChatGPT-assisted virtual patient scenarios? 3. What do you think about the contribution of the Chat GPT-assisted virtual patient application to communication skills?The probing questions 4. What are your suggestions for enhancing empathy in the ChatGPT-assisted virtual patient application? 5. If audio and video features were added to the virtual patient, would anything change in terms of empathy? 6. So, while I can use my (communication) skills this much with a real patient, can you say that I used them better or worse with this patient? Clarification question 7. Yes, so you’ve become more willing to maintain communication with the virtual patient, am I understanding correctly? (Focus group 3) 8. Do you think time is an advantage in this practice? Do you think there’s another advantage to this? In terms of communication skills? (Focus group 4) Summarizing and clarifying 9. Well, based on your statements, this was easier for you because the virtual patient in front of you communicated more openly with you. As far as I understand, we agree on this. (Focus group 1) 10. Concluding11. Do you have any thoughts on the impact of the virtual patient interview experience on communication and empathy skills?

Focus group interviews were conducted and lasted 25–30 min. Table 2 presents the schedule and participant groupings. Discussions addressed two primary topics: communication and experiences with empathy. Facilitators adhered to a predetermined list of questions and requested additional details as necessary to obtain comprehensive responses. The interviews were recorded and transcribed using Microsoft Word’s dictation tool. The accuracy of the transcripts was verified by simultaneously listening to the audio recording during transcription.

Table 2.

Focus group sessions and participants’ demographic characteristics

Focus Group Sessions	Date	Number of Participants	Demographic characteristics
First Focus Group	November 27, 2024	4 students (NS* 1, 2, 3, and 4)	NS 1: Female, 22 years old NS 2: Male, 24 years old NS 3: Male, 22 years old NS 4: Male, 22 years old NS 5: Male, 21 years old NS 6: Male, 21 years old NS 7: Female, 20 years old NS 8: Male, 20 years old NS 9: Male, 20 years old NS 10: Male, 24 years old NS 11: Male, 21 years old NS 12: Female, 21 years old NS 13: Female, 20 years old NS 14: Female, 20 years old NS 15: Male, 20 years old
Second Focus Group	December 4, 2024	3 students (NS 5, 6, and 7)
Third Focus Group	December 11, 2024	4 students (NS 8, 9, 10, and 11)
Fourth Focus Group	December 18, 2024	4 students (NS 12, 13, 14 and 15)

* Nursing Students

Within the focus groups, the primary researcher led the discussion while the secondary researcher posed follow-up and clarifying questions. Both researchers were faculty members and study investigators with experience leading group sessions. This dual role may have influenced students due to potential power dynamics. To mitigate this risk, participation was voluntary, discussions occurred in a relaxed setting outside the classroom, and the moderator encouraged open sharing. An additional the first researcher (A.K.D) observed the sessions to minimize the moderator’s influence further.

Data saturation

Across the four focus groups and the accompanying structured diaries, recurring patterns were consistently observed, and no substantially new themes emerged in the later datasets. Participants consistently described the contribution of the ChatGPT-supported virtual patient interaction to communication and empathy awareness, future communication strategies, and professional preparation. Similarly, recurring challenges, including difficulty conveying sincerity and obtaining detailed patient information through appropriate questioning, were reported across both data sources. These recurring patterns indicated that no substantially new concepts emerged in the later datasets.

Data analysis

The data were analyzed in Microsoft Word 11 using the thematic analysis approach [15]. First, we created separate transcripts for each focus group and each journal to capture detailed, individualized data. Next, both the focus group and structured journal transcripts were compiled in Word files, organized according to the responses given to each question. The coding process commenced following the completion of the interventions and the conclusion of data collection. The first author assigned superscript codes to meaningful text units to support systematic coding, and the corresponding codes were listed. Then, the code list was shared with the second author, who examined whether the codes accurately represented the participants’ statements. In cases of disagreement, the authors discussed and decided which code best reflected the meaning of the statement. We generated a list of codes, which a second researcher reviewed to check that they matched the participants’ statements. Finally, we listed all codes in Microsoft Word and grouped them into themes based on their relationships. Themes and their codes were checked for coherence and relevance. All Word analysis files were archived under separate headings for systematic organization.

Trustworthiness

Lincoln and Guba’s four trustworthiness criteria were followed [16]. To ensure credibility, focus group interviews and reflective journals were used within data triangulation. Data triangulation was performed by comparing the findings from focus group interviews with the structured reflective journals completed by participants after each focus group session. The two datasets were analyzed separately during the initial coding phase. When inconsistencies were identified, researchers re-examined the original transcripts and journal entries, assessed the context of the relevant data, and discussed alternative interpretations until consensus was reached. Codes from the two separate datasets were sorted to form meaningful groups. This process was performed for all focus group interview transcripts, and finally, all codes were combined.

Transferability was supported by providing a detailed description of the study setting, participant characteristics, sampling strategy, and data collection procedures. Regarding dependability, the data collection and analysis process was described in detail. All research data and analysis results were preserved for independent evaluation. For confirmability, coding was conducted and verified independently by two researchers.

Results

This study identified three main themes; Theme 1: “Unrealistic empathy and communication experience”, Theme 2: “Solution-oriented and effective communication experience”, Theme 3: “Communication and empathy awareness: Future communication strategies and professional preparation”. Theme 1 highlights perceived limitations in the experience of communication and empathy. These limitations include the perceived lack of realism in empathy and ChatGPT’s failure to reflect realistic communication challenges. In contrast, Theme 2 describes how ChatGPT’s interaction style facilitated solution-oriented communication, increased confidence, and reduced communication-related stress despite these limitations. Building on these experiences, Theme 3 reflects participants’ awareness of effective communication and empathy, the development of future communication strategies, and perceived preparation for clinical practice. Together, these themes demonstrate that although the virtual patient interaction was not perceived as fully realistic, it provided a supportive learning environment that promoted reflection and preparation for future patient communication.

Theme 1: Unrealistic empathy and communication experience

Participants reported that although they were able to empathize with the virtual patient, they did not experience this empathy as realistic. Maintaining communication through writing, without gestures, facial expressions, or tone of voice, hinders comprehensive understanding of the patient’s concerns and emotional state. This situation makes it difficult for students to understand the patient’s experiences, condition, and problems through written virtual patient scenarios alone.

“Not using body language, especially since I already know the person, I am corresponding to is an artificial intelligence, prevents me from empathizing (…). If it were in front of me and I could see it, its use of body language and its constant questioning would have made the situation more beneficial for me.” (NS6).

“I could not really empathise with ChatGPT. Once more, there was plenty of empathy, but I do not think it was a genuine patient experience. We tend to read patients better from their facial expressions, gestures, or tone of voice. Virtual patients says it only has a problem because of this, but I am not sure to what extent it affects them. (…)” (NS14).

Participants appear to have developed new approaches to experiencing empathy more effectively. Participants stated that while chatting with ChatGPT, they empathized by conceptualizing the virtual patient as a real patient and imagining themselves in the other person’s position.

“(…) The virtual patients had encountered misinformation on social media and came to me for counseling. In terms of empathy, I literally saw myself in the patient’s shoes. I could understand his concerns and understand that many of his questions were perfectly normal.” (NS9).

In addition, ChatGPT did not fully reflect the complexity of real-world communication. Participants mentioned that they were unable to use their skills in managing difficult patients and making rapid decisions because the virtual patient was easy to persuade, showed no resistant behaviors, and offered a limited range of realistic emotional responses. In real-world practice, patients may exhibit hesitation, skepticism, emotional distress, or resistance; this requires healthcare providers to employ more sophisticated communication strategies. The lack of realistic interaction reduced authenticity and limited opportunities for data collection, rapid decision-making, and practicing managing skeptical or questioning patients.

“This simulation helped me communicate with the virtual patient. I was able to communicate more easily, but I do not think ChatGPT provided enough feedback because it was convincing right away. However, compared to real life, patients are not easily convinced. (…).” (NS9).

Overall, participants recommended developing more realistic scenarios, integrating emotional expressions, and adding audio and video features to ChatGPT. The role of sound and image features, as well as emotional expressions, in facilitating empathy was a particular point of emphasis among the participants. The suggestions are based on a more realistic formulation of ChatGPT. The need for alternative scenarios arises from the diversification of virtual patient profiles. It was emphasized that profiles such as patients with low socio-cultural levels and anxious or angry relatives of patients, as well as realistic patient feedback, should be incorporated into the scenarios.

“We can also enter profiles, for example. A patient may have multiple solutions. Someone may ask questions. For example, it (the ChatGPT) may be nervous, anxious, or worried about its illness. (…) When we encounter that patient at the clinic, we can apply the treatment we have entered in their profile.” (NS3).

“A system where patients’ emotions are more clearly indicated, or ChatGPT’s voice feature, could provide a more realistic experience.” (NS8).

Theme 2: Solution-oriented and effective communication experience

Despite the unrealistic communication and empathy experience, participants experienced effective communication, solution-oriented communication, self-confidence, relaxed interaction, and reduced stress during the ChatGPT interaction.

ChatGPT’s active participation in the virtual patient role, and its provision of clear, focused, and systematic responses, enabled students to communicate effectively. Participants reported that by asking appropriate questions, they were able to better understand the problem and offer potential solutions.

“First, when I think about the patient’s concerns one by one and respond accordingly, it is a more comprehensive answer, and what I have to say is not confusing. When we conduct training in the hospital, things can sometimes get confusing. However, in this virtual scenario, I was able to express myself more clearly, for example, by communicating more easily. It had a significant impact on my communication skills. I get the feeling that I am sharing more useful info somewhere. For example, when I listened to the patient’s problems and gave them answers, it improved my communication skills. (…)” (NS11).

Overall, this solution-oriented communication experience fostered happiness, a sense of accomplishment, and self-confidence in students. Participants reported that they were able to direct the flow of the interview in the scenario and comfortably express their knowledge. This provided a sense of accomplishment and individual motivation.

“(…) I chatted to ChatGPT, and it seemed like he could communicate with me well and that my ideas could work for him. It made me feel like I had really achieved something.” (NS2).

“Finding solutions to the client’s questions and problems and making suggestions made me feel both happy and like a professional. I’ve never really talked to anyone about these issues in such detail and guided them before.” (NS3) The same participant found being consulted as knowledgeable, motivating.

“Honestly, being consulted as someone with expertise motivates me. Finding answers for them and helping them regain their health is a positive thing for me.” (NS3).

Participants attributed these positive experiences to ChatGPT’s interface design and interaction style, which created a comfortable communication environment by reducing performance pressure and allowing time to reflect before responding. This helps participants communicate more confidently, reduce stress, and practice their communication and empathy skills.

“As a student, I get nervous when I am with a real patient, which makes it hard for me to give the right information and to empathize. But since we have the chance to speak in a more relaxed and understandable way in the face of this simulation and have an idea about how to communicate, (…)” (NS3, In structured journal).

“(The virtual patient) has drawn a profile that can easily tell its troubles and problems (…). It is very good at explaining the problems it’s facing. We can get back what we’ve taught to a high extent. So, because it was solution-oriented, we identified the problems well and made a return accordingly.” (NS2).

Theme 3: Communication and empathy awareness: future communication strategies and professional preparation

The lessons students learned from their ChatGPT communication experience included awareness, reflection, communication planning, empathy awareness, and future communication strategies. They explained that their ChatGPT experiences fostered an awareness of the need for a holistic approach that respects the physical, psychosocial, and emotional well-being of patients, and described the communication strategies they developed as a result of this awareness. These strategies focus more on empathy, and asking necessary questions to gather more detailed data from patients.

“Whenever I’d give the patient new info, he’d ask a new question, which made me think more. It was really important to me because I realized that I needed to think about things that my real patients might be hesitant to ask or say.” (NS6).

“(…) I realized I needed to take a more holistic approach, thinking about not only the physical problems patients experience but also their mental and emotional health. (…) In the future, I’ll make sure to really listen to patients and pay attention to what they’re saying. If you give patients time to talk without rushing them, it’ll help build trust and make for better communication. (…)” (NS1, Structured journal).

“When I was contacting my patient (Chat GPT) (…) I learned that I need to ask more detailed questions. I think that this will make me try even harder to improve my ability to conduct focused interviews with my patients by thinking more carefully about the questions I ask, and that I will be more eager to engage in dialogue.” (NS4).

In addition, this practice serves as preliminary training for internships and professional preparation, and signaled that this experience would be advantageous in future ventures.

“I think this simulation was very useful for us. It created a stepping stone between us and real patients. So, if we encounter this in our future internships, we can keep this in mind and provide better training (to the patient).” (NS10).

“Yeah, it’s a great way to solidify what we know. Because when we’re with a real patient, we can be like, “Oh yeah, I remember this part from the simulation…” So there are common points. Yes, these are common questions… I’ll definitely remember this part. So I’ll have some experience, and it’ll help us feel more comfortable when we’re with a patient.” (NS8).

“This experience is going to be really beneficial for my professional development. First, it’ll help me reinforce my knowledge and make it more memorable. Second, I think it’ll help me to keep my cool in a situation like that. Third, it’ll help me see where I’m not doing so well and work on that, which will help me grow as a person.” (NS5, Structured journal).

Discussion

This study identified nursing students’ experiences of ChatGPT-supported virtual patient communication. These findings revealed that nursing students perceived ChatGPT-supported virtual patient communication as an unrealistic communication and empathy experience; however, ChatGPT’s interaction structure offered a systematic, solution-oriented, and comfortable communication environment. Furthermore, this experience contributed to students developing reflections on communication and empathy, recognizing their strengths and areas for improvement, shaping their future communication strategies, and supporting their professional preparation for clinical practice.

The lack of realism participants experienced in empathizing with the virtual patient — driven by the absence of nonverbal cues such as gestures, facial expressions, and tone of voice — is consistent with prior research on this limitation. Previous studies have also observed this absence of nonverbal communication. For example, in a qualitative study, medical students found ChatGPT suitable and accessible for patient communication. However, they felt that the lack of nonverbal cues, emotional tone, and authentic interpersonal interaction limited empathy. Another study reported that some students had difficulty developing empathy during interactions with ChatGPT due to the nature of written communication [17]. Furthermore, while virtual patients can contribute to empathy training, they cannot replace interactions with real patients [18]. Collectively, these findings highlight that text-based AI virtual patient applications lack nonverbal communication. As a result, they are insufficient on their own to recreate the authenticity of empathic clinical encounters. Despite this, participants appeared to compensate through cognitive empathy strategies, such as imagining the virtual patient as a real person to better understand their concerns. This aligns with findings that students mainly showed empathy by attempting to understand the patient’s symptoms, asking appropriate questions, and using clinical reasoning [19], even when face-to-face elements were limited.

The study also identified several limitations in ChatGPT’s ability to fully capture the complexity of real-world clinical communication. The absence of challenging communication behaviors may have restricted opportunities to develop advanced communication skills required for managing complex patient encounters. Previous studies have noted that large-scale language models can alter the virtual patient’s role, provide unrealistic responses, and deviate from the scenario, failing to fully reflect the resistance behaviors of real patients or the emotional and social complexity of clinical communication [13, 19]. It is also noted that digital or virtual communication environments sometimes fail to fully reflect the social, emotional, and relational aspects of face-to-face communication. For this reason, virtual dialogues may sometimes be perceived as more artificial, limited, and transactional [20]. Therefore, given these limitations, AI-based simulations should be viewed as complementary educational tools for experiencing communication and empathy skills.

Regarding empathy and communication experiences that felt unrealistic, participants offered suggestions. These suggestions include integrating audio and video, adding emotional expressions, using different patient profiles that exhibit angry or highly intense emotional reactions, and challenging patient scenarios. The literature emphasizes the need for more realistic AI simulations. For example, Jeon and Jeong [11] argued that integrating voice, facial expressions, and virtual human technologies could significantly enhance the realism of AI-based patient simulations. It has been reported that students participating in a more realistic generative AI-based virtual patient system, SimBot, gained greater confidence in communication and found the simulation realistic [12]. Chen and Li [21] showed that ChatGPT simulations with audio and visual features create a safe and immersive learning environment for nursing students. Real-time feedback and realistic simulations significantly enhance communication confidence and clinical skills [22]. Therefore, future developments in AI-supported educational technologies should focus on integrating multimodal communication features to better mimic real clinical interactions.

The another key fingings, ChatGPT’s interaction style offers a relaxed and user-friendly environment with clear and problem-focused interaction. Participants perceived the clear and systematic interaction as facilitating problem identification and appropriate responses, while also enhancing self-confidence, a sense of accomplishment, and individual motivation. These findings are in line with various studies in the literature. It is stated that virtual patient scenarios provide students with repeatable, low-risk, and structured communication opportunities, which increase self-confidence and communication competence [4, 23]. In addition, ChatGPT’s provision of a comfortable communication environment has a positive impact on the communication experience. This positive impact is particularly the reduced pressure to respond quickly or perfectly, and performance anxiety. This finding is consistent with simulation-based learning theory, which emphasizes the importance of psychologically safe environments for skill development. A separate study similarly reported that AI-supported simulations can enhance self-efficacy and learning motivation by allowing students to experiment with communication strategies without the fear of harming real patients [10]. Research on AI-based patient simulations has noted similar benefits: Cross, Kayalackakom [19] found that such simulations provide students with flexible, repeatable, and self-paced learning, while Kim and Won [12] noted that students improved their skills in asking questions, gathering data, and communicating with patients in a lower-stress environment. Therefore, ChatGPT-supported simulations may serve as a supportive educational tool for developing communication skills.

Although participants perceived the interaction as lacking the realism of actual patient encounters, this limitation did not hinder certain gains. On the contrary, the structured, low-pressure communication environment fostered reflection by allowing students to practice asking questions, organize their clinical reasoning, and identify areas requiring further development before interacting with real patients. These findings suggest that the educational value of ChatGPT-assisted virtual patient communication lies not in precisely replicating real clinical encounters, but in providing a psychologically safe environment for communication practice and reflective learning.

Building on these experiences, the findings suggest that ChatGPT-supported virtual patient communication fostered students’ awareness of effective communication and empathy, encouraging reflection and the development of future communication strategies. Similarly, Andrade [24] highlighted that virtual patient technologies could support the development of professional competencies by enabling pharmacy students to engage in repeated patient interactions before real-world clinical experience. Jeon and Jeong [11] found that through AI-based simulation, students developed strategies such as using short and simple sentences, speaking slowly, and rephrasing explanations when communicating with elderly patients. This has enabled students to learn new communication techniques and prepare them for real-world clinical situations [11]. Lesińska-Sawicka and Michalak [25] noted that AI simulations can facilitate students’ self-assessment and gain insights from their experiences by encouraging them to re-evaluate their assumptions and communication approaches while interacting with patients. By engaging students in interactive dialogue, virtual patients can encourage students to critically evaluate their communication strategies and develop greater awareness of patient-centered care principles.

Taken together, these findings suggest that although ChatGPT-supported virtual patient communication cannot replace authentic patient encounters, it can complement communication education by promoting reflection, communication awareness, and preparation for subsequent clinical learning experiences.

Limitations

This study has several limitations. First, participants were recruited from a single nursing faculty using a qualitative sample; therefore, the findings may not be transferable to other educational settings. Second, although participation was voluntary and confidentiality was assured, socially desirable responses cannot be excluded. Finally, one researcher was a research assistant at the same faculty and the other was the course instructor, which may have influenced participants’ responses because of the existing educational relationship. To minimize this potential bias, participants were informed that participation was voluntary, their responses would remain confidential, and their decision to participate or the content of their responses would not affect their course evaluation or academic standing. This study examined students’ perceived experiences rather than their actual implementation of empathetic communication skills or the effectiveness of the intervention; future research employing observational or performance-based assessment methods is needed to evaluate this.

Conclusion

Overall, this study’s key finding is that nursing students experienced ChatGPT-supported virtual patient simulations as a useful, low-risk opportunity to practice communication and reflect on their empathy skills, even though the empathy and communication they experienced did not feel entirely realistic. Despite this limitation, participants reported that the structured, low-pressure nature of the interaction reduced performance anxiety and encouraged a more holistic approach to patient communication. Participants also identified specific ways the simulation could be made more realistic, such as incorporating audio-visual and emotional features. These findings suggest that, in its current text-based form, ChatGPT-supported virtual patient simulation was perceived by students as a complementary tool that supports reflection on communication and empathy, although whether it effectively prepares students for subsequent clinical communication remains to be evaluated in future research.

Author contributions

A.K.D: Writing-original draft, Investigation, Supervision, Conceptualization, Methodology, Formal analysis, Data curation, Prepeared figure 1.A.D.A: Writing – review & editing, Investigation, Supervision, Conceptualization, Resources. All authors have read and agreed to the published version of the manuscript.

Funding

None.

Data availability

Although not included in the article or in the Supplementary Information (for raw data, not summary data such as mean and variance), all interview transcripts and written journal data are available on the researcher’s computer for reuse or review.

Declarations

Ethics approval and consent to participate

Ethical approval was granted by the Akdeniz University Clinical Research Ethics Committee (TBAEK-740). All participants were informed, agreed to participate, and were identified by the code “NS” (Nursing Students). The study was conducted in accordance with the ethical principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Generative AI and AI-assisted technologies in the writing process

During the preparation of this work the authors used DeepL Translator and Grammarly in order to translation from Turkish to English and editing of the translation. Furthermore, Chat GPT is employed in research procedures and the writing process to enhance readability. After using this tools, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

Competing interests

The authors declare no competing interests.