Italian Nursing Students Say It Is Too Early to Judge AI Chatbots in Their Education

Artificial intelligence chatbots have swept through universities faster than most curricula can adapt, and one of the groups standing at the front line of that disruption is nursing students — the future professionals who will translate digital tools into bedside care. A new cross-sectional multicentre study from Italy, published in BMC Medical Education, offers one of the most detailed snapshots yet of how this cohort actually knows, uses, and judges AI chatbots, and the results reveal a striking paradox: nearly everyone has heard of these tools, but more than half say it is simply too early to have a clear opinion about whether they belong in nursing education.

The research team, led by Fabio Tortella of the Azienda Ospedaliera Universitaria Integrata di Verona and Silvia Gianola of the IRCCS Istituto Ortopedico Galeazzi in Milan, with senior authors Giacomo Rossettini and Alvisa Palese, surveyed students enrolled in nursing programs at three Italian universities. The questionnaire was developed by a panel of experts based on the available literature and pilot-tested before deployment, and the survey itself was reported following three established methodological checklists: the Checklist for Reporting Results of Internet E-Surveys (CHERRIES), the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE), and the Consensus-Based Checklist for Reporting of Survey Studies (CROSS). That triple-reporting discipline matters, because online e-surveys are notoriously vulnerable to inflated response rates and self-selection bias, and the authors anchored their findings in a framework designed to keep those distortions in check.

In total, 812 nursing students participated, a response rate of 58.4 percent — a respectable figure for an online academic survey. The headline finding is one of near-universal awareness paired with cautious ambivalence. Most participants were aware of AI chatbots, with 700 of the 812 respondents, or 86.2 percent, reporting that they had heard of them. Yet when asked to rate the usefulness of these tools, the median response was only 2 out of 4, with an interquartile range of 2 to 3 — a rating the researchers characterized as “Somewhat useful.” And when the survey turned to the introduction of AI chatbots in education specifically, 445 students, or 54.8 percent, said it was too early to have a clear opinion. In other words, the generation that grew up with smartphones is not rushing to endorse the technology that could reshape how they learn to care for patients.

Beneath the aggregate numbers, the inferential statistics exposed demographic fault lines that the authors say carry real educational implications. Increasing age was associated with lower odds of having heard of AI chatbots, with an odds ratio of 0.96 (95 percent CI 0.93 to 0.99; p = 0.021), and with lower odds of having used them for academic purposes, with an odds ratio of 0.97 (95 percent CI 0.94 to 0.99; p = 0.046). The effect sizes are small per year of age, but across a cohort spanning multiple years they compound into measurable awareness gaps. Male students had higher odds of being aware of AI chatbots than female students, with an odds ratio of 2.06 (95 percent CI 1.19 to 3.55; p = 0.009), and among those with previous academic use, male students also had higher odds of reporting more frequent use, with an odds ratio of 2.15 (95 percent CI 1.28 to 3.63; p = 0.004). The gender gap in both awareness and intensity of use suggests that the benefits of AI tools may not be distributed evenly across the future nursing workforce, which is predominantly female.

Perhaps the most counterintuitive finding concerns academic progression. Students with regular academic progression had lower odds of previous academic use of AI chatbots than those experiencing difficulties in their academic progression, with an odds ratio of 0.71 (95 percent CI 0.51 to 0.99; p = 0.044). The authors interpret this pattern as evidence that students facing academic challenges may be turning to chatbots as compensatory aids — tools that help them keep pace when conventional study strategies fall short. That interpretation raises a double-edged question for educators: are struggling students using AI to genuinely support their learning, or are they leaning on a technology that masks underlying gaps in knowledge and competence? The study cannot answer that question directly, but it flags the population where educational vigilance may be most needed.

Digital self-confidence emerged as another powerful predictor. Students who rated their own digital competence positively had higher odds of reporting a greater likelihood of future AI chatbot use, with an odds ratio of 1.97 (95 percent CI 1.08 to 3.59; p = 0.026). This finding cuts to the heart of the digital literacy debate: perceived competence, not just actual competence, shapes whether students will engage with emerging technologies. If a student believes they are digitally capable, they are roughly twice as likely to say they will use AI chatbots going forward. The authors argue that this self-perception is precisely the kind of variable that educational interventions can target — building both genuine skills and the confidence to apply them responsibly.

Technically, the study’s design reflects the current state of AI chatbot research in health professions education. The term “AI chatbot” in this context refers to conversational systems built on large language models — the technology behind generative pre-trained transformers (GPT) — which produce fluent, human-like text in response to natural language prompts. These systems can summarize clinical guidelines, explain pharmacological mechanisms, draft care plans, and answer study questions, but they also carry well-documented limitations: they can generate plausible-sounding but incorrect information, they lack clinical judgment, and they cannot replace the relational core of nursing practice. The authors’ expert-panel questionnaire was designed to capture students’ knowledge of these tools, their patterns of academic use, and their perceived benefits and limitations, providing a structured evidence base rather than anecdote.

The methodological rigor of the survey deserves emphasis in an era when AI adoption studies often rely on convenience samples. By recruiting across three universities and reporting against CHERRIES, STROBE, and CROSS simultaneously, the team addressed the reproducibility crisis that has plagued e-survey research. The response rate of 58.4 percent, while not perfect, is substantially higher than the single-digit rates common in unsolicited online surveys, and the multicentre design reduces the risk that findings reflect the culture of a single institution. The use of descriptive statistics such as the interquartile range alongside inferential measures such as odds ratios with 95 percent confidence intervals gives readers a transparent view of both central tendency and uncertainty.

What do these findings mean for the future of nursing education? The authors conclude that the evidence supports the inclusion of digital literacy education and ethical guidance in nursing programs to promote informed and responsible use of AI chatbots. The near-universal awareness combined with the majority’s hesitation about educational integration suggests that students are not resisting the technology — they are waiting for frameworks that tell them when, how, and why to use it. The gender and age gaps identified in the study point to populations where targeted interventions may be most valuable, and the association between academic difficulty and AI use highlights the need for educators to understand what struggling students are actually doing with these tools.

The study also arrives at a moment when health systems worldwide are grappling with how to prepare the next generation of clinicians for an AI-augmented workplace. Nursing is the largest profession in health care, and the students surveyed here will spend their careers alongside systems that increasingly embed conversational AI — from triage support to patient education platforms. The Italian findings suggest that the pipeline from awareness to competent, ethical use is not automatic. It requires deliberate curriculum design, explicit ethical guidance, and continued research into the implications of AI for learning outcomes. The authors call for further research to evaluate the implications of AI chatbots for nursing education and to identify appropriate strategies for their integration into the curriculum. Until those strategies exist, the students’ own verdict stands: the technology is here, everyone knows about it, but it is too early to say what it will do to the way nurses learn to care.

Subject of Research: Medical Education

Article Title: What are the knowledge, use, and perceived benefits and limitations of artificial intelligence chatbots among Italian nursing students? A cross-sectional multicentre study

Article References: Tortella, F., Gianola, S., Castellini, G., Turolla, A., Landuzzi, M. G., Cook, C., Corradi, F., Bulfone, G., Canzan, F., Chiappinotto, S., Sist, L., Rossettini, G., & Palese, A. (2026). What are the knowledge, use, and perceived benefits and limitations of artificial intelligence chatbots among Italian nursing students? A cross-sectional multicentre study. BMC Medical Education. https://doi.org/10.1186/s12909-026-10439-9

Image Credits: AI Generated

DOI: 10.1186/s12909-026-10439-9

Keywords: artificial intelligence, AI chatbots, nursing education, nursing students, digital literacy, medical education, Italy, survey research, knowledge, perceived, benefits, limitations