Integrating ChatGPT into nursing education: a randomized trial exploring knowledge, retention, and student perspectives on endotracheal suctioning Authors: Emine Sezgunsay, Ahmet Berkay Polat, Sude Bengisu Kılıcer Published: 2026 ABSTRACT AimThis study aimed to evaluate the preliminary effectiveness of a ChatGPT-integrated educational session on nursing students’ knowledge acquisition and retention regarding endotracheal suctioning, and to explore their perspectives and learning experiences.MethodsA pilot pre-test/post-test parallel-group randomized controlled trial with an embedded qualitative component was conducted with 32 undergraduate nursing students who participated voluntarily. Participants were randomly assigned to either a ChatGPT-integrated education group or a traditional education group. Knowledge was assessed using a structured test administered before, immediately after, and six weeks following the intervention. Quantitative data were analyzed using appropriate non-parametric statistical tests, while qualitative data were analyzed thematically following Braun and Clarke’s approach.ResultsWhile no significant difference was observed between groups in immediate post-test knowledge scores, the ChatGPT-integrated group demonstrated significantly higher knowledge retention over time at the six-week follow-up (p < 0.05). Qualitative findings revealed three main themes: learning supportive factors, student gains, and suggestions for improvement. Participants reported that interactive learning supported by ChatGPT facilitated active participation and reinforced learning.ConclusionsThe findings suggest that ChatGPT-integrated education has the potential to enhance knowledge retention over time, although it may not significantly impact immediate learning outcomes. Given small sample size, these result should be considered preliminary. Future large-scale studies are recommended to examine long-term outcomes and address potential implementation challenges.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12909-026-09483-2. Introduction Chat Generative Pre-trained Transformer (ChatGPT) is an artificial intelligence (AI) language model that integrates natural language processing techniques and large language models with generative AI, possessing the capability to comprehend complex linguistic patterns. Owing to these features, it can generate coherent, human-like texts, provide diverse responses to queries, deliver answers based on predefined scenarios, create interactive experiences by engaging with users, and offer contextually robust outputs on a variety of topics [34]. Due to these potentials, ChatGPT has rapidly gained a broad user base [38] and, similar to other sectors, has attracted the attention of students, educators, and researchers in the field of nursing education [3, 15, 32]. The integration of ChatGPT into educational settings holds the potential to reshape learning experiences by providing students with personalized and interactive platforms [20]. Students can utilize ChatGPT to create personalized study notes, review complex concepts, ask questions about study materials and receive instant responses, as well as benefit from its support in preparing assignments and presentations [32]. In this context, ChatGPT is perceived as a versatile tool that enhances the effectiveness of learning experiences. However, the phenomenon known as “hallucination,” in which the system produces highly fluent yet inaccurate or misleading information, poses the risk of directing students toward erroneous knowledge [9, 10]. Moreover, its inability to access password-protected scientific databases represents a limitation in reaching up-to-date and evidence-based information [10]. Therefore, it is crucial that students develop the ability to critically evaluate the information provided by AI chatbots, particularly in determining whether it offers reliable knowledge on a given subject [28, 40]. On the other hand, the use of ChatGPT in students’ assignment preparation processes has sparked significant debates regarding ethical concerns in AI applications, plagiarism, and academic integrity [36, 29, 33]. Nursing educators and researchers can utilize ChatGPT for curriculum planning, the development of clinical case assignments, the creation of constructivist or collaborative learning environments within courses, the design of assessment drafts, or the analysis of data [1]. Moreover, attempts have been made to integrate ChatGPT not merely as a traditional text-based response software but into virtual or face-to-face laboratory simulations, where it has been reported to enhance student engagement through interactive learning experiences [29, 36]. Nevertheless, the risks associated with the use of ChatGPT by nursing students similarly apply to nursing educators. Educators who integrate ChatGPT into course content should implement reference-based and detailed verification processes in order to ensure the accuracy and reliability of the generated information [16]. Furthermore, concerns have been raised that excessive reliance on ChatGPT may undermine students’ abilities in independent thinking, critical analysis, problem-solving, and innovative approaches [40]. While utilizing ChatGPT as a tool to enhance the effectiveness of their teaching, educators should also instruct their students on the safe, effective, and ethical use of this AI technology. In the future, healthcare systems and care management mechanisms that take advantage of such features are expected to be developed, and nurses must be prepared to work with these systems. Despite the rapid adoption of AI applications in healthcare and the need for adequate preparation to enable healthcare providers to practically implement these technologies, there is still limited evidence on how ChatGPT and similar AI tools can be integrated into education, how students’ critical thinking and clinical reasoning competencies can be ensured, and how such tools can be used ethically [15]. Therefore, to better understand the impact of ChatGPT on nursing education, further empirical research is needed to examine how it influences students’ knowledge and skill acquisition as well as their engagement in learning. Background The broad accessibility and multilingual support of ChatGPT enable easier access to AI technologies even for individuals with limited technical infrastructure or financial resources [40]. Nursing educators, in turn, aim to benefit from this accessibility within their own professional domains while also evaluating the impact of such technologies on nursing education. From a pedagogical perspective, the integration of ChatGPT represents a shift toward more student-centered and inquiry-based learning environments. Accordingly, leveraging the strengths and opportunities offered by ChatGPT is of considerable importance for its effective integration into nursing education [1]. However, although the body of literature at the intersection of ChatGPT and nursing education is steadily growing, the majority of existing publications consist of editorials, commentaries, and opinion papers [20, 21, 37, 43]. Existing review studies have primarily focused on identifying the strengths and weaknesses of ChatGPT [1], employing text network analysis to define thematic clusters rather than examining its effectiveness in use [21], or analyzing bibliometric trends [24, 47]. While these reviews highlight the transformative potential of ChatGPT in nursing education, they also report that the literature is predominantly composed of non-experimental publications [3, 32] and that methodological heterogeneity across studies poses challenges for comparing results [23]. A review of the literature reveals studies in which nursing educators have integrated ChatGPT into their courses and reported the outcomes. Hsu [25] employed a chatbot based on crossword puzzles and ChatGPT to support nursing students in learning medical terminology and examined its effectiveness. Chang and colleagues reported that courses incorporating a mobile chatbot enhanced students’ academic performance, self-efficacy, critical thinking skills, and learning satisfaction [13, 13, 14, 14]. Arkan et al. [5] found that ChatGPT-based training not only improved students’ problem-solving skills and fostered positive attitudes toward artificial intelligence, but also contributed to enhancing their competence in the nursing process and increasing overall satisfaction [5]. Tseng et al. [42], approaching from a different perspective, noted that AI-assisted reports prepared by students lacked professional and individualized nursing interventions, and therefore recommended the guided use of AI tools such as ChatGPT or Copilot under teacher supervision [42]. More recently, Gümüş and İlknur [22] reported that a ChatGPT-integrated intercultural nursing course created an inclusive and reflective learning environment that enhanced students’ cultural competence and had a positive impact on their communication skills [22]. Despite these positive reports, a critical gap remains: most studies focus on general academic performance or soft skills, leaving the impact of AI on specialized, high-stakes clinical procedures—such as endotracheal suctioning—largely unexplored. As the use of ChatGPT and other AI tools in nursing education expands, the importance of academic research in this area is also increasing [28]. To ensure the safe and effective integration of AI into nursing curricula, there is a need to design diverse educational interventions and examine their impact on students and learning outcomes. Such efforts will not only support educators and nursing students in the appropriate use of current AI applications but also enable them to take the lead in shaping future implementations [18]. However, before large-scale implementation, it is crucial to conduct pilot studies that evaluate how ChatGPT can be integrated not as a standalone source of information, but as a supplementary instructional tool within established pedagogical frameworks. To the best of our knowledge, no study has yet investigated the impact of a ChatGPT-integrated endotracheal suctioning session —structured as a supplementary instructional tool rather than a primary information source— on nursing students’ knowledge and experiences. Addressing this gap, the present pilot study aimed to provide preliminary, hypothesis-generating evidence on the effects of ChatGPT-integrated endotracheal suctioning session on undergraduate nursing students’ knowledge acquisition, knowledge retention, and perceptions of their learning experiences. Furthermore, it seeks to extend the literature on the integration of AI into nursing education by evaluating the potential of ChatGPT as a dynamic teaching material and by proposing practical recommendations for future large-scale implementations. Methods Aim The aim of this pilot study was to evaluate the preliminary effect of a ChatGPT-integrated educational session on nursing students’ knowledge acquisition and retention regarding endotracheal suctioning procedures, and to explore their perspectives and experiences through qualitative methods. The hypotheses were: I.There is a difference between the intervention and control groups in terms of knowledge scores regarding endotracheal suctioning.II.Nursing students who receive a ChatGPT-integrated education session demonstrate higher knowledge retention compared to those receiving traditional education. I.There is a difference between the intervention and control groups in terms of knowledge scores regarding endotracheal suctioning. There is a difference between the intervention and control groups in terms of knowledge scores regarding endotracheal suctioning. II.Nursing students who receive a ChatGPT-integrated education session demonstrate higher knowledge retention compared to those receiving traditional education. Nursing students who receive a ChatGPT-integrated education session demonstrate higher knowledge retention compared to those receiving traditional education. Study design This study was designed as a pilot pre-test/post-test parallel-group randomized controlled trial with an embedded qualitative component. To ensure ethical standards and maintain the principle of voluntary participation, the study was conducted with a convenience sample of students who volunteered outside of their regular curriculum hours. Nursing students were randomly assigned to either the intervention group, which received a ChatGPT-integrated educational session, or the control group, which received a traditional education session, using a computer-generated randomization list. To minimize contamination between groups, the control and intervention sessions were conducted in consecutive weeks. Quantitative data were collected at baseline, immediately after the intervention, and six weeks later to assess knowledge retention. In addition, qualitative data regarding students’ perspectives and experiences were collected through open-ended questions and analyzed thematically. The study was conducted in accordance with the Consolidated Standards of Reporting Trials (CONSORT) guidelines (see Fig. 1). The Clinical Trial Number NCT07096518 has been retrospectively registered at ClinicalTrials.gov (https://clinicaltrials.gov/study/NCT07096518) on 7 August 2025, in accordance with ethical and clinical research protocols.Fig. 1Consolidated standards of reporting trials (CONSORT) Consolidated standards of reporting trials (CONSORT) Participants, sample size, and setting The population of the study consisted of 44 fourth-year and 48 third-year nursing students enrolled in the spring semester of the 2024–2025 academic year at the Department of Nursing of a private university in Turkey. The inclusion criteria were: (1) being a native-level speaker of Turkish (in terms of speaking and comprehension), and (2) having successfully passed the Fundamentals of Nursing course taken during the first year of the program. Initially, 34 students agreed to participate in the study; however, two students were excluded due to non-attendance at the scheduled educational sessions. As a result, the final sample consisted of 32 students (24 fourth-year and 8 third-year). A post-hoc power analysis was conducted based on the results of the Mann–Whitney U test comparing the retention scores between the intervention and control groups. Using the observed effect size (r = 0.356), a total sample size of 32 (17 intervention, 15 control), and a significance level of α = 0.05, the achieved power was calculated as 0.63, indicating a moderate level of statistical power. Randomization The random allocation sequence was generated by the researchers. Participants were first stratified by year of study (third- and fourth-year), and then by grade point average (GPA ≤ 2.0 and > 2.0). Random assignment was subsequently conducted within each stratum using the Research Randomizer tool (https://www.randomizer.org/). This stratified randomization approach ensured a balanced distribution of participants in terms of academic level and performance across groups. At the beginning of the study, 34 students were randomly assigned in a 1:1 ratio to the intervention (n = 17) and control (n = 17) groups. To ensure allocation concealment, the generated sequence was kept in an opaque, password-protected electronic file accessible only to the researcher responsible for group assignments, ensuring that the enrollment of participants was independent of the knowledge of their future group allocation. Blinding and bias control Due to the nature of the educational intervention, blinding of participants and the instructor was not feasible. To minimize potential bias, all educational sessions were conducted by the same instructor using standardized teaching materials, including PowerPoint presentations and whiteboard explanations. To address the risk of contamination, students were explicitly requested not to share the content or methods of their respective sessions with peers until the entire study was completed. Additionally, sessions were scheduled in consecutive weeks to reduce the duration of potential interaction between groups. Data collection tools Student information form The form begins with three questions inquiring about the student's pseudonym, age, and gender. It continues with two questions regarding the use of ChatGPT for course preparation and prior experience with ChatGPT-based education. The final section includes three questions assessing whether the student has previously performed endotracheal suctioning in a clinical setting, their willingness to perform the procedure, and their perceived competence. The form consists of a total of eight questions and was administered to the students prior to the educational sessions. The Suctioning Knowledge Questionnaire (SKQ) The SKQ was developed in Turkish by Özden and Yılmaz in 2021 for use with nursing students [48]. The questionnaire includes 20 items that assess knowledge of endotracheal suctioning procedures across three phases: before, during, and after the intervention. Each item is multiple-choice with five response options. Correct responses are awarded five points, and incorrect responses receive zero points, resulting in a total possible score ranging from 0 to 100. The Content Validity Index of the questionnaire was 0.98. Reliability analysis using the split-half method yielded a Spearman-Brown coefficient of 0.781, indicating a high level of internal consistency. In the present study, the SKQ was administered to the students as a pre-test before the educational sessions, as a post-test after the sessions, and again six weeks later as a follow-up test to assess knowledge retention. Student feedback form This form was developed by the researchers to gather qualitative feedback from students following the educational sessions. It consists of 6 open-ended questions aimed at exploring students’ thoughts and feelings about the instructional process and materials. The form was administered to the students at the end of the educational sessions. Study process Educational procedures for the control and intervention groups The sessions conducted with both the control and intervention groups were not part of any official course listed in the academic curriculum. Therefore, the day and time of the lessons were selected based on the students’ availability to ensure optimal participation. The control group received their lesson first, followed by the intervention group in the following week. Each educational session lasted for two 45-min class periods. Control group procedures Before the lesson, students in the control group were informed about the study, and written informed consent was obtained from those who voluntarily agreed to participate. Subsequently, they completed the Student Information Form and the SKQ, which was administered as a pre-test. The lesson on endotracheal suctioning was delivered using traditional teaching methods, including PowerPoint presentations, discussions, and question-and-answer sessions. At the end of the educational session, endotracheal suctioning procedures were demonstrated in the classroom using a mannequin, a portable aspirator, and endotracheal suction materials, and students who wished were given the opportunity to practice the procedure themselves. At the end of the session, students were asked to retake the SKQ as a post-test and fill out the " Student Feedback Form." Intervention group procedures Similar to the control group, students in the intervention group were first informed about the study and provided informed consent. They then completed the Student Information Form and the SKQ as a pre-test. During the two 45-min class periods, education was delivered through a ChatGPT-integrated approach. ChatGPT version 3.5 was used during the educational session. Students were provided with guidance on how to interact with ChatGPT, including information on effective prompt usage to optimize their learning experience. Students followed a step-by-step learning process, including question exploration, peer discussion, and comparison with clinical guidelines, while using ChatGPT. To ensure consistency, students were guided in the use of standardized prompts during their interaction with ChatGPT (Appandix 1). Each student was assigned a question related to endotracheal suctioning, which was developed in parallel with the main topics of the AARC Clinical Practice Guidelines: Artificial Airway Suctioning [7]. Students were instructed to use ChatGPT to find answers to their assigned questions and then discuss their findings with a peer. Following this, the Turkish-translated version of the AARC Clinical Practice Guidelines was shared with the students, who were asked to locate the relevant sections in the guideline and compare the information with the responses generated by ChatGPT. After all groups completed their tasks, the answers were shared collectively in class, followed by a question-and-answer session. The instructor actively monitored the process,if ChatGPT generated incomplete or incorrect information (hallucinations), students were encouraged to identify these discrepancies during the comparison phase with the AARC Clinical Practice Guidelines. The instructor then provided immediate corrective feedback during the collective classroom discussion to ensure that all participants acquired accurate clinical knowledge. As in the control group, at the end of the educational session the students in the intervention group also observed a demonstration of the endotracheal suctioning procedure and were given the opportunity to practice the procedure themselves. At the end of the session, students were again asked to retake the SKQ as a post-test and fill out the feedback form. Both groups received the same core educational content, differing only in the instructional approach. Knowledge retention assessment To evaluate knowledge retention, the SKQ was re-administered to all participants six weeks after the intervention. Since the academic semester had ended and most students had returned to their homes, the test was converted into a digital format to ensure accessibility and feasibility of data collection. The digital version of the knowledge test form was distributed to the students electronically. Participants were instructed to complete the test independently and submit the completed digital form within the specified time frame. Participants were instructed to complete the test independently without external assistance. Data analysis The research data were analyzed using IBM SPSS Statistics for Windows, Version 21.0 (IBM Corp., Armonk, NY, USA). Descriptive statistics, including frequency, percentage, minimum and maximum values, median, and standard deviation, were used to summarize the data. Since the data did not meet the assumptions of normal distribution, and the study was designed as an exploratory pilot trial, non-parametric tests were applied. The Mann–Whitney U test was used to compare differences between the control and intervention groups at each independent time point. Changes within groups across measurement points (pre-test, post-test, and follow-up) were analyzed using pairwise comparisons with the Wilcoxon signed-rank test. Although a repeated-measures design is typically preferred for longitudinal analysis, point-by-point non-parametric comparisons were utilized to maintain statistical conservativeness given the limited sample size. Statistical significance was set at p < 0.05. The qualitative data were analyzed using the six-phase thematic analysis framework described by Braun and Clarke, which involved: (1) familiarization with the data, (2) generating initial codes, (3) searching for themes, (4) reviewing themes, (5) defining and naming themes, and (6) producing the report [8]. Recorded responses to the open-ended questions were transcribed verbatim into Microsoft Word. To ensure familiarity with the data, all researchers read and re-read the transcripts. Initial codes were generated independently by each researcher through open coding, capturing meaningful units of text relevant to the research objectives. Codes were then organized into potential themes and subthemes, which were refined and finalized through collaborative discussion until consensus was reached. An audit trail documenting coding decisions was maintained to enhance the credibility and transparency of the process. All transcripts were anonymized to protect participants’ confidentiality. Reflexivity was maintained throughout the analysis by researchers engaging in continuous dialogue to identify and mitigate potential personal biases. Initial codes were generated independently, and any discrepancies were resolved through a consensus-based approach, ensuring that the final themes accurately represented the participants' collective feedback. Ethical considerations Ethical approval for this study was granted by the Izmir University of Economics Health Sciences Research Ethics Committee on 22 April 2025 (Approval No: E-97429853–050.04- 95,745). All procedures performed in this study involving human participants were conducted in accordance with the ethical standards of the institutional research committee and with the Declaration of Helsinki. Additionally, institutional permission was obtained from the faculty where the research was conducted, allowing the study to take place within the faculty’s classrooms. Prior to participation, students were informed about the objectives and procedures of the study, including their right to withdraw at any time without facing any consequences. Written informed consent was obtained from all students who agreed to take part. Authorization to use the SKQ was received from the original developers. All students enrolled in the third and fourth years of the nursing program were invited to participate in the study. Before data collection began, students were informed about the study's purpose, procedures, and their rights, including the option to withdraw at any time without any negative consequences. They were explicitly assured that participation was voluntary and would not affect their academic progress or evaluations. The educational intervention and data collection were conducted outside of the students’ scheduled curriculum hours and were not part of a compulsory course session. The knowledge test, included as part of the research, was used solely for data collection purposes and was not considered in students’ academic grading or performance evaluation. Written informed consent was obtained from all participants prior to the commencement of the study. Results A total of 32 students completed the study, with 17 in the intervention group and 15 in the control group (Table 1). All participants in the intervention group were female (100%, n = 17), while the control group included one male student. In both groups, the majority of students reported using ChatGPT to prepare for their coursework; however, only a few had previously received any formal education using the ChatGPT artificial intelligence tool. All participants expressed a desire to experience endotracheal suctioning procedures in clinical practice. The majority of students in both groups stated that they had not previously performed endotracheal suctioning on real patients and did not feel competent to independently perform the procedure in a clinical setting. No significant differences were found between the groups in terms of age, gender, prior ChatGPT use, or experiences with endotracheal suctioning (p > 0.05).Table 1Descriptive characteristics of the participants (n = 32)Categories/QuestionsIntervention group(n = 17)Control group(n = 15)pMean ± SDMean ± SDAge21.73 ± 1.43822.40 ± 1.1210.097an/%n/%GenderWoman17/10014/93.30.279bMan-1/6.7Do you use the ChatGPT to prepare for your courses?”Yes16/94.114/93.30.927bNo1/5.91/6.7Have you previously received any education using the ChatGPT?Yes4/23.53/201.000bNo13/76.512/80Have you experienced performing endotracheal suctioning in clinical practice?Yes11/64.77/46.70.503bNo6/35.38/53.3Do you consider yourself competent to independently perform endotracheal suctioning on a real patient?Yes3/17.61/6.70.603bNo14/82.414/93.3perform endotracheal suctioning in a clinical practice setting?Yes17/10015/100-a Mann–Whitney U testb Ci-Square test/Fisher’s Exact Test* p < 0,05 Table 1Descriptive characteristics of the participants (n = 32)Categories/QuestionsIntervention group(n = 17)Control group(n = 15)pMean ± SDMean ± SDAge21.73 ± 1.43822.40 ± 1.1210.097an/%n/%GenderWoman17/10014/93.30.279bMan-1/6.7Do you use the ChatGPT to prepare for your courses?”Yes16/94.114/93.30.927bNo1/5.91/6.7Have you previously received any education using the ChatGPT?Yes4/23.53/201.000bNo13/76.512/80Have you experienced performing endotracheal suctioning in clinical practice?Yes11/64.77/46.70.503bNo6/35.38/53.3Do you consider yourself competent to independently perform endotracheal suctioning on a real patient?Yes3/17.61/6.70.603bNo14/82.414/93.3perform endotracheal suctioning in a clinical practice setting?Yes17/10015/100-a Mann–Whitney U testb Ci-Square test/Fisher’s Exact Test* p < 0,05 Descriptive characteristics of the participants (n = 32) a Mann–Whitney U test b Ci-Square test/Fisher’s Exact Test * p < 0,05 Knowledge acquisition When comparing the knowledge levels measured before the educational session, there was no statistically significant difference between the intervention and control groups (Z = −0.687, p = 0.492). At the end of the educational session, both groups demonstrated a statistically significant increase in their knowledge scores compared to their respective pre-test results (Table 2). However, no statistically significant difference was found between the post-test scores of the intervention and control groups (Z = −1.260, p = 0.208).Table 2Comparisons of pre-test and post-test knowledge scores in the control and intervention groups (n = 32)GroupPre-testPost-testPTest ValueMedian (IQR)(Min–Max)Median (IQR)(Min–Max)Intervention group (n = 17)55(20–75)70(45–85)0.003*−2.960aControl group (n = 15)45(30–75)75(55–85)0.001*−3.423aP0.492b0.208bTest Value−0.687−1.260Bold text is used to highlight statistically significant resultsa Wilcoxon testb Mann–Whitney U test* p < 0,05 Table 2Comparisons of pre-test and post-test knowledge scores in the control and intervention groups (n = 32)GroupPre-testPost-testPTest ValueMedian (IQR)(Min–Max)Median (IQR)(Min–Max)Intervention group (n = 17)55(20–75)70(45–85)0.003*−2.960aControl group (n = 15)45(30–75)75(55–85)0.001*−3.423aP0.492b0.208bTest Value−0.687−1.260Bold text is used to highlight statistically significant resultsa Wilcoxon testb Mann–Whitney U test* p < 0,05 Comparisons of pre-test and post-test knowledge scores in the control and intervention groups (n = 32) Median (IQR) (Min–Max) Median (IQR) (Min–Max) 55 (20–75) 70 (45–85) 45 (30–75) 75 (55–85) Bold text is used to highlight statistically significant results a Wilcoxon test b Mann–Whitney U test * p < 0,05 Knowledge retention When the post-test and retention test scores of the students were compared, a statistically significant decrease was observed in the knowledge scores of both groups six weeks after the educational session (Table 3). However, when the retention scores were compared between the groups, the intervention group had significantly higher knowledge scores than the control group, indicating better knowledge retention among students who received the intervention (Z = −2.014, p = 0.044).Table 3Comparisons of post-test and retention test knowledge scores in the control and intervention groups (n = 32)GroupPost-testFollow-Up TestPTest ValueMedian (IQR)(Min–Max)Median (IQR)(Min–Max)Intervention group (n = 17)70(45–85)65(25–80).011*−2.547aControl group (n = 15)75(55–85)55(15–75).001*−3.423aP0.208b0.044bTest Value−1.260−2.014Bold text is used to highlight statistically significant resultsa Wilcoxon testb Mann Whitney U test* p < 0,05 Table 3Comparisons of post-test and retention test knowledge scores in the control and intervention groups (n = 32)GroupPost-testFollow-Up TestPTest ValueMedian (IQR)(Min–Max)Median (IQR)(Min–Max)Intervention group (n = 17)70(45–85)65(25–80).011*−2.547aControl group (n = 15)75(55–85)55(15–75).001*−3.423aP0.208b0.044bTest Value−1.260−2.014Bold text is used to highlight statistically significant resultsa Wilcoxon testb Mann Whitney U test* p < 0,05 Comparisons of post-test and retention test knowledge scores in the control and intervention groups (n = 32) Median (IQR) (Min–Max) Median (IQR) (Min–Max) 70 (45–85) 65 (25–80) 75 (55–85) 55 (15–75) Bold text is used to highlight statistically significant results a Wilcoxon test b Mann Whitney U test * p < 0,05 Qualitative findings on educational experience Participants’ feedback on their educational session experience was analyzed using Braun and Clarke’s thematic analysis method. Three main themes emerged: learning supportive factors (opportunity for hands-on training, use of materials, interactive learning environment, scope of educational content), student gains (knowledge acquisition, increased awareness, recall and reinforcement of information, retention), and suggestions for improvement (various recommendations regarding the implementation of the session and the educational materials) (Table 4).Table 4Themes, subthemes, and representative student quotesThemesSubthemesStudents’ quotesLearning Supportive FactorsInteractive learning environment*IG-P23: I felt sufficiently active. Distributing a question to everyone and discussing it helped us gain better mastery of the topicIG-P29: It was interactive; researching the questions ourselves first contributed to my learning, which was greatIG-P31: Sufficient activity was ensured by everyone. Our own research on the topic further enhanced this**CG-P5: Being interactive and able to immediately practice on the mannequinCG-P2: I think I was active because I practiced one-on-oneCG-P10: I was quite active. The number of participants allowed everyone to be actively involvedActive participation and ownershipIG-P29: The aspect that most supported my learning was researching the topic independentlyIG-P28: Although I was not accustomed to presenting answers to questions individually after researching them myself, I found this method effectiveIG-P27: Using artificial intelligence and explaining what we learned to each otherIG-P24: Receiving questions, researching them ourselves, and observing the differencesStudent gainsIncreased awarenessIG-P30: I thought I knew how to perform suctioning, but by the end of this educational session, I realized I had forgotten some essential information. Now I have recalled itIG-P24: I noticed that something we often see being done is frequently performed incorrectly. The practical component of the educational session was particularly importantIG-P23: Discussing the correct and incorrect ways to perform a procedure we may encounter throughout our professional life increased our level of awarenessRetentionIG-P17: Conducting our own research helped me retain the topic in more detailIG-P30: Because we progressed interactively, it was more lastingIG-P15: The participation of attendees made the educational session effective and memorableCG-P9: I think practicing contributed to increased retentionSuggestions for improvementSuggestionsIG-P30: We could have performed the final mannequin practice before this educational session to see what had changed and what could help us learn betterIG-P26: Experiences, case studies, and examples could also have been included in the lessonIG-P22: It was quite good. The same question could be given to both ChatGPT and the instructor, allowing for a discussion between the twoCG-P7: More visuals and videos could be added to the presentation content*IG: Intervention group**CG: Control group Table 4Themes, subthemes, and representative student quotesThemesSubthemesStudents’ quotesLearning Supportive FactorsInteractive learning environment*IG-P23: I felt sufficiently active. Distributing a question to everyone and discussing it helped us gain better mastery of the topicIG-P29: It was interactive; researching the questions ourselves first contributed to my learning, which was greatIG-P31: Sufficient activity was ensured by everyone. Our own research on the topic further enhanced this**CG-P5: Being interactive and able to immediately practice on the mannequinCG-P2: I think I was active because I practiced one-on-oneCG-P10: I was quite active. The number of participants allowed everyone to be actively involvedActive participation and ownershipIG-P29: The aspect that most supported my learning was researching the topic independentlyIG-P28: Although I was not accustomed to presenting answers to questions individually after researching them myself, I found this method effectiveIG-P27: Using artificial intelligence and explaining what we learned to each otherIG-P24: Receiving questions, researching them ourselves, and observing the differencesStudent gainsIncreased awarenessIG-P30: I thought I knew how to perform suctioning, but by the end of this educational session, I realized I had forgotten some essential information. Now I have recalled itIG-P24: I noticed that something we often see being done is frequently performed incorrectly. The practical component of the educational session was particularly importantIG-P23: Discussing the correct and incorrect ways to perform a procedure we may encounter throughout our professional life increased our level of awarenessRetentionIG-P17: Conducting our own research helped me retain the topic in more detailIG-P30: Because we progressed interactively, it was more lastingIG-P15: The participation of attendees made the educational session effective and memorableCG-P9: I think practicing contributed to increased retentionSuggestions for improvementSuggestionsIG-P30: We could have performed the final mannequin practice before this educational session to see what had changed and what could help us learn betterIG-P26: Experiences, case studies, and examples could also have been included in the lessonIG-P22: It was quite good. The same question could be given to both ChatGPT and the instructor, allowing for a discussion between the twoCG-P7: More visuals and videos could be added to the presentation content*IG: Intervention group**CG: Control group Themes, subthemes, and representative student quotes *IG-P23: I felt sufficiently active. Distributing a question to everyone and discussing it helped us gain better mastery of the topic IG-P29: It was interactive; researching the questions ourselves first contributed to my learning, which was great IG-P31: Sufficient activity was ensured by everyone. Our own research on the topic further enhanced this **CG-P5: Being interactive and able to immediately practice on the mannequin CG-P2: I think I was active because I practiced one-on-one CG-P10: I was quite active. The number of participants allowed everyone to be actively involved IG-P29: The aspect that most supported my learning was researching the topic independently IG-P28: Although I was not accustomed to presenting answers to questions individually after researching them myself, I found this method effective IG-P27: Using artificial intelligence and explaining what we learned to each other IG-P24: Receiving questions, researching them ourselves, and observing the differences IG-P30: I thought I knew how to perform suctioning, but by the end of this educational session, I realized I had forgotten some essential information. Now I have recalled it IG-P24: I noticed that something we often see being done is frequently performed incorrectly. The practical component of the educational session was particularly important IG-P23: Discussing the correct and incorrect ways to perform a procedure we may encounter throughout our professional life increased our level of awareness IG-P17: Conducting our own research helped me retain the topic in more detail IG-P30: Because we progressed interactively, it was more lasting IG-P15: The participation of attendees made the educational session effective and memorable CG-P9: I think practicing contributed to increased retention IG-P30: We could have performed the final mannequin practice before this educational session to see what had changed and what could help us learn better IG-P26: Experiences, case studies, and examples could also have been included in the lesson IG-P22: It was quite good. The same question could be given to both ChatGPT and the instructor, allowing for a discussion between the two CG-P7: More visuals and videos could be added to the presentation content *IG: Intervention group **CG: Control group Learning supportive factors Students’ feedback highlighted that certain aspects of the education provided to both the intervention and control groups were perceived as strengths and supported their learning processes. These supportive elements were categorized into four subthemes: opportunities for hands-on practice, use of materials, interactive learning environments, and the scope of the educational content. Opportunities for hands-on practice: Regarding the hands-on practice, students in both groups emphasized the practical sessions on endotracheal suctioning performed on mannequin at the end of the theoretical instruction. They reported that the opportunity to apply what they had learned reinforced their understanding and represented a strong aspect of the educational session. Some examples of students’ responses are presented below. Intervention group (IG)-P16: “The theoretical educational session is better understood when combined with practice.” Intervention group (IG)-P16: “The theoretical educational session is better understood when combined with practice.” Control group (CG)-P1: “Reinforcing it through practice was a strong point of the education I received.” Control group (CG)-P1: “Reinforcing it through practice was a strong point of the education I received.” Use of materials: Students in both groups provided positive feedback regarding the presentations used in the course, the materials employed during the endotracheal suctioning educational session, and the effectiveness of ChatGPT in creating interactive content. They indicated that the use of these materials throughout the course supported their learning. IG-P21: “The educational session was supported with materials, presentations, and hands-on practice.” IG-P21: “The educational session was supported with materials, presentations, and hands-on practice.” CG-P12: “The presentations allowed me to adequately understand the necessary information.” CG-P12: “The presentations allowed me to adequately understand the necessary information.” Interactive learning environment: Analysis of feedback from both groups indicated that students perceived the course as interactive. However, the scope of interactivity differed between the groups, with this emphasis being more pronounced in the intervention group. Students in the control group considered the instructor posing questions during the presentation and experiencing the endotracheal suctioning procedure as aspects of interactivity. In contrast, students in the intervention group interpreted interactivity as responding to questions using ChatGPT, sharing answers with peers, ensuring everyone had the opportunity to speak, participating in critical thinking discussions conducted via the question-and-answer method, and experiencing the endotracheal suctioning procedure. Both groups also reported that the small class size facilitated active participation. IG-P31: “Sufficient activity was ensured by everyone. Our research on the topic also increased this engagement.” IG-P31: “Sufficient activity was ensured by everyone. Our research on the topic also increased this engagement.” CG-P2: “I believe I was active because I practiced one-on-one.” CG-P2: “I believe I was active because I practiced one-on-one.” Scope of educational content: Students in both groups emphasized that the educational content was concise, clear, and comprehensive. The adequacy of the course duration and content, along with the clarity of instruction, was reported to be a source of satisfaction among the students. IG-P31: “The presentation was concise, clear, and included the necessary information; having the opportunity to observe it during practice and our own research supported my learning.” IG-P31: “The presentation was concise, clear, and included the necessary information; having the opportunity to observe it during practice and our own research supported my learning.” CG-P6: “The topics were presented concisely and without information overload. The completeness and adequacy of the materials were strong aspects of the education.” CG-P6: “The topics were presented concisely and without information overload. The completeness and adequacy of the materials were strong aspects of the education.” Student gains Feedback from students in both the intervention and control groups highlighted the learning outcomes provided by the educational session on endotracheal suctioning. These outcomes were analyzed under the subthemes of knowledge acquisition, awareness, and retention. Knowledge acquisition: All participating students emphasized that their knowledge regarding endotracheal suctioning increased. The opportunity to practice at the end of the course allowed them to apply this theoretical knowledge as a skill, providing significant learning gains. Additionally, students noted that although they had previously completed courses on endotracheal suctioning, they could not recall certain information, and this educational session offered them the chance to revisit and reinforce their knowledge. The small class sizes in both groups also appeared to enhance students’ motivation, further supporting knowledge acquisition. IG-P19: “Learning the principles of suctioning and the small number of participants supported my learning.” IG-P19: “Learning the principles of suctioning and the small number of participants supported my learning.” CG-P2: “It was effective in helping me understand how much pressure to apply during suctioning, how many times to repeat it, and the correct way to insert and remove the catheter.” CG-P2: “It was effective in helping me understand how much pressure to apply during suctioning, how many times to repeat it, and the correct way to insert and remove the catheter.” Awareness This theme emerged only in the feedback from students in the intervention group. The students reported that they recognized gaps in their prior knowledge and developed awareness regarding the correct execution of a procedure they are likely to encounter frequently in professional settings, particularly in intensive care environments. IG-P30: “I thought I knew how to perform suctioning, but by the end of this educational session, I realized that I could not recall some of the information I needed. I was able to remember it.” IG-P30: “I thought I knew how to perform suctioning, but by the end of this educational session, I realized that I could not recall some of the information I needed. I was able to remember it.” Retention Students in the intervention group particularly emphasized that the educational method enhanced the retention of the knowledge they acquired. In the control group, only one student reported that practicing suctioning contributed to the retention of the learned material. Most students in the intervention group indicated that, in addition to the skills practice, using ChatGPT in class to search for answers to questions, sharing responses with peers, and engaging in class-wide question-and-answer discussions helped consolidate and retain their learning. IG-P17: “Conducting my own research helped me retain the topic in more detail in my mind.” IG-P17: “Conducting my own research helped me retain the topic in more detail in my mind.” CG-P9: “I believe it contributed to increasing retention.” CG-P9: “I believe it contributed to increasing retention.” Suggestions for improvement When examining this theme, it is observed that students in the intervention group provided more structured responses compared to those in the control group. Students in the control group suggested using more visuals in the presentation and including videos. Students in the intervention group, in addition to the suggestions made by the control group, provided further constructive recommendations, such as allowing students to perform suctioning on a mannequin before the educational session and again after the educational session to observe differences, increasing case discussions or experience sharing related to suctioning, giving the same question to two students before the large group discussion so they could discuss it among themselves, and creating more questions to facilitate class discussions. Discussion This study examined the effects of ChatGPT-integrated education on nursing students’ knowledge acquisition and retention in endotracheal suctioning. The findings demonstrated that while ChatGPT-integrated education did not result in significantly higher immediate knowledge scores compared to traditional teaching methods, it was associated with improved knowledge retention over time within this specific study group. Importantly, this quantitative finding was supported by structured student feedback, as students in the ChatGPT-integrated group reported that the learning activities were perceived to enhanced the retention of their knowledge and increased their awareness of the procedure. In addition, participants described feeling more active and engaged during the learning process. Knowledge acquisition and retention The findings of our study demonstrated that both traditional education and ChatGPT-integrated session were effective in improving nursing students’ knowledge of endotracheal suctioning. No statistically significant difference was found between the two groups in post-educational knowledge assessments. However, it is essential to frame these findings as preliminary evidence and hypothesis-generating evidence, as the small sample size and exploratory nature of the study may affect the stability of the quantitative results. These findings suggest that the integration of ChatGPT may support deeper cognitive processing rather than short-term memorization. While both instructional approaches were effective in improving immediate knowledge levels, the superior retention observed in the ChatGPT group indicates that interactive and student-centered learning processes may contribute to more durable learning outcomes. In the literature, there are studies examining the effectiveness of ChatGPT and similar AI-based software engines in enhancing nursing students’ academic achievement. Research investigating the use of AI-based chatbots among nursing students has reported positive outcomes: Lee, Hwang and Chen [27] found improved academic performance in public health courses,Hsu [25] reported greater success in learning medical terminology among those using AI-based chatbots, [13, 14] demonstrated increased academic performance in physical examination courses, and Chang Hwang and Gau (2022) reported improved knowledge of obstetric immunization. In contrast to some previous studies reporting improvements in immediate academic performance with AI-based tools, our findings did not demonstrate a clear advantage in post-test knowledge scores. This difference may be explained by variations in instructional design, duration of exposure, and the way AI tools are integrated into the learning process. The integration of ChatGPT into the nursing education as a dynamic learning assistant may simplify students’ access to core resources and enhance their ability to acquire information in real time [1]. In our intervention, the educational mechanism relied on students using ChatGPT as a starting point for inquiry, followed by verification against clinical guidelines. This process is crucial because educators must be cautious about AI-generated "hallucinations"—the dissemination of inaccurate data or factual errors. The literature highlights that to address these challenges, ChatGPT-generated content should be carefully reviewed before use, and educators should provide guidance until students develop the necessary skills to critically evaluate such information [1, 22, 31]. Furthermore, the educational session provided on the use of ChatGPT has been shown to significantly improve nurses’ and nursing students’ attitudes toward AI [5, 26]. Our study demonstrates the feasibility of testing such an intervention in a controlled classroom setting, though the controlled nature of the trial may limit its generalizability to real-world clinical environments. Our findings showed that knowledge levels regarding endotracheal suctioning decreased in both groups six weeks after the educational session, although they remained above baseline. Importantly, students in the ChatGPT-integrated group demonstrated significantly higher retention scores compared to those in the traditional education group, indicating that AI integration may support knowledge retention. While promising, this observation is based on a single comparison with a relatively small effect size and should be treated as a foundation for further high-powered trials rather than a definitive conclusion. The structured use of ChatGPT, combined with peer discussion and comparison with clinical guidelines, may have facilitated active engagement and deeper understanding, which in turn supported long-term knowledge retention. Studies examining the effects of AI-based tools on knowledge retention are still limited. Başaran and Duru [6] investigated the effects of sexual health education delivered to nursing students using Kahoot, ChatGPT, and traditional methods. Similar to our findings, their study showed that knowledge levels declined by the third month but did not drop below baseline. Unlike our results, however, Başaran and Duru [6] reported that ChatGPT supported knowledge acquisition less effectively than other methods. Nevertheless, the way ChatGPT was integrated into the education in their study was not clearly described. Therefore, well-structured and larger-scale studies are needed to examine the effects of ChatGPT and other AI-based chatbots on knowledge retention. Perspectives and opinions The primary aim of the qualitative analysis of the feedback collected in this study was to explore students’ perceptions and experiences regarding the integration of ChatGPT into the course. The qualitative insights obtained here reflect structured student feedback on a novel intervention rather than an exhaustive exploration of the student experience. However, participants evaluated the educational session in a more holistic manner, which contributed to the emergence of a broader conceptual framework. The first theme identified was related to factors that support learning. Students emphasized that the opportunity for hands-on training and the use of supportive materials were strong aspects of the session. It is well established that providing nursing students with opportunities to practice at their own pace in safe environments—whether in laboratories, well-designed clinical practice settings, or through simulations at varying levels of fidelity—supports not only their knowledge and skill acquisition but also the development of their professional identity [4, 12, 30]. Most of the students who participated in our study reported that the scope of the educational sessions was sufficient and met their expectations. The participants were 3rd- and 4th-year nursing students who had previously completed several internships and observed many nursing practices in clinical settings. Therefore, they are considered individuals capable of identifying their own shortcomings and learning needs. In addition, the students who volunteered for this study were those who already had questions about endotracheal suctioning and were motivated to improve their skills in this area. This was evident both from their voluntary participation and from their responses to the student information form. Accordingly, the fact that the students stated they found answers to their questions during the educattional session suggests the effectiveness of both the theoretical sessions and the open and closed endotracheal suctioning practices conducted at the end of the educational session. As another subtheme reported under the factors supporting learning, the concept of an interactive learning environment emerged more strongly in the feedback of the intervention group. Within this feedback, students emphasized that they felt more active and more confident in participating when the educational session was conducted in small groups. This finding is consistent with previous studies exploring the effectiveness of nursing education, which have shown that small group learning is considered more effective [11, 46]. However, the notable distinction in our study regarding the concept of interactivity lies in the way it was defined by the two groups: while the control group described interactivity as having opportunities to perform the practice and being asked questions by the instructor during educational session, the intervention group defined interactivity as finding answers to questions about endotracheal ssuctioning with the assistance of ChatGPT, discussing these answers collectively in class, and then practicing the procedure. Thus, the interactivity described by the intervention group reflects a more comprehensive understanding aligned with the principles of constructivist learning. In constructivist education, students are placed at the center of the learning process by employing critical thinking, collaboration, and personal inquiry skills, while the instructor’s role is to facilitate learning within the group by encouraging participation and providing appropriate feedback [17, 35]. Consistent with this, existing studies have reported that ChatGPT and other AI-powered language models foster active participation in learning among nursing students [13, 14, 25]. By integrating ChatGPT into classroom activities, simulations, or discussion forums, educators can employ it as an interactive element that captures students’ interest and promotes active learning [1]. Within the theme of student learning outcomes, participants reported that the educational session refreshed the knowledge related to endotracheal suctioning that they had previously acquired in earlier courses. While students converged under the subtheme of knowledge acquisition, they diverged in the themes of awareness and retention, with the intervention group providing extensive feedback particularly in these areas. This appears to be linked to the classroom discussions supported by ChatGPT. During these discussions, every student participated; in explaining answers to their peers, they occasionally drew upon their own clinical observations, while at other times they directly posed questions to one another. These exchanges fostered greater awareness of the importance of properly carrying out endotracheal suctioning in clinical practice. Feedback from the intervention group further indicated that independently seeking answers to questions and then presenting these findings to peers in class enhanced knowledge retention. The study’s quantitative findings also support this perspective. Although previous research has examined the retention of nursing education, many studies emphasize the need for further investigation to identify factors influencing retention [49, 19, 39, 41]. Given that the present study was conducted with a small sample and retention was assessed only once, further research is required to illuminate the potential effects of AI tools on knowledge retention in nursing education. The qualitative insights obtained here should be viewed as structured student feedback on a novel educational intervention rather than a comprehensive exploration of the student experience. The suggestions collected from students upon completion of the study also provide important insights for structuring nursing education. Students expressed a desire to watch videos demonstrating how procedures are performed during practical training, as merely explaining the steps of a procedure in theory—even if they had observed it in clinical settings—was not sufficient for them to mentally visualize the practice. Their need to observe critical points of the procedure learned in theoretical sessions was interpreted as a natural and legitimate requirement. Furthermore, student feedback indicated a preference for more intensive question-based discussions, case analyses, and interactive methods such as performing procedures before educational session and then discussing the mistakes observed. These responses were interpreted as a reflection of students’ desire to position themselves at the center of the learning process. In such interactive approaches, where students learn through engagement with peers, they are allowed to develop their own professional skills and behaviors. Interactive teaching methods are grounded in Vygotsky’s sociocultural learning and social constructivist theories, and their effectiveness in nursing education continues to be explored [2, 45, 44]. Limitations One of the limitations of this study is that it was conducted at a single university with a small sample size, which may leave the trial underpowered for detecting subtle differences. In addition, approximately 65% of eligible nursing students declined participation or did not provide informed consent. Since the study was conducted outside regular class hours, students’ academic workload and time constraints may have negatively affected participation. Consequently, the study sample consisted only of students who voluntarily chose to participate, suggesting that participants may systematically differ from non-participants in terms of interest in the research topic, motivation, or time management skills. Therefore, self-selection bias may have arisen, potentially limiting both the internal and external validity of the study results. Given these constraints, this study should be viewed as an exploratory pilot trial and the results as hypothesis-generating rather than definitive. Due to the nature of the intervention, blinding of participants and the instructor was not feasible, which poses a risk of performance and observer bias. Furthermore, despite scheduling sessions in consecutive weeks, the risk of peer contamination between groups cannot be entirely excluded. Participants’ awareness of group allocation may have influenced their motivation or engagement, while the instructor’s unintentional attitudes could have affected the delivery of the intervention. Additionally, the repeated use of the same assessment tool (SKQ) may have triggered a testing effect, influencing post-test gains through recall. While the six-week follow-up was beneficial, its remote and unsupervised nature introduces potential variability in testing conditions and the risk of external assistance. Furthermore, the possibility of students engaging in independent study or continuing to use AI tools during the six-week gap could not be fully controlled. Finally, the absence of systematic participation data restricted a comprehensive assessment of the students’ level of engagement with the educational content. Conclusion The research findings indicate that the integration of AI-based chatbots, such as ChatGPT, into nursing education has the potential to positive effects on student learning outcomes particularly in enhancing knowledge retention. Incorporating AI applications into the curriculum could not only support the development of students’ knowledge but also enhance the retention of learned information. Students were found to exhibit positive attitudes toward the integration of these technologies into educational processes. Moreover, the use of AI-based tools in the classroom may help students gain awareness and experience regarding the ethical and responsible use of such technologies. In conclusion, while AI integration shows promise in improving the quality of nursing education, the results of this exploratory trial suggest it should be viewed as a supplementary tool rather than a replacement for traditional methods at this stage. Given the pilot nature and the small sample size of this study, these conclusions should be considered preliminary and interpreted with caution. Future studies should focus on evaluating the role of AI tools in enhancing nursing education and clinical preparedness, with an emphasis on conducting rigorous randomized controlled trials to ensure reliable evidence. Examining the effectiveness of applications designed to support the safe use of AI tools in improving students’ or nurses’ knowledge and skills is also essential. In this process, exploring the influence of personal factors—such as individual characteristics and attitudes toward AI—on learning experiences may provide more realistic and applicable insights. Addressing the use of AI in learning from multiple perspectives will allow for a deeper understanding of the dynamics that affect students’ performance and perceptions, thereby enabling researchers and educators to design more effective learning approaches in the future. Supplementary Information Supplementary Material 1. Supplementary Material 1.