The study, led by Wenyu Cheng and colleagues at the First Hospital of Shanxi Medical University in Taiyuan, was published in BMC Nursing. Rather than treating AI literacy as a single score that rises or falls from nurse to nurse, the team used a statistical technique called latent profile analysis, which searches for hidden subgroups within a population whose responses share a common pattern. This person-centered approach is designed to capture heterogeneity that averages conceal, and it revealed a workforce far more stratified than conventional surveys suggest.

Artificial intelligence literacy, as the researchers frame it, is a multidimensional construct. It encompasses awareness of what AI systems can and cannot do, the practical ability to apply them, the capacity to evaluate their outputs critically, and an understanding of the ethical questions that arise when algorithms touch patient care. A nurse might score high on one dimension and low on another, and the combination matters: a clinician who can operate a predictive early-warning dashboard but cannot question its recommendations occupies a very different professional position from one who combines hands-on fluency with ethical vigilance.

When the latent profile analysis was run, four patterns emerged. The first and smallest group was labeled the actively embracing type, nurses who scored consistently high across awareness, application, evaluation, and ethics. The second, the knowledge-application dissociation type, describes nurses who understand AI conceptually but fail to translate that knowledge into practice. The third, the skill-oriented type, shows the reverse configuration: solid procedural competence paired with comparatively weaker grasp of underlying knowledge and critical evaluation. The fourth and most concerning group, the weak literacy type, scores low across the board. Together, the middle two categories accounted for the majority of the 1,160 participants, while the actively embracing type represented the smallest share of the sample.

The technical distinction between these profiles is more than academic taxonomy. In a hospital setting, the knowledge-application dissociation profile implies a training pipeline that successfully delivers theory, through lectures, policy documents, or awareness campaigns, but stops short of supervised, hands-on engagement with clinical AI tools. The skill-oriented profile suggests the opposite failure mode: nurses learn button-level operation of specific systems without developing the conceptual foundation that would let them generalize to new tools, spot errors, or reason about algorithmic bias. Both patterns represent half-built literacy, and each demands a different corrective intervention.

To understand what pushes nurses into one profile or another, the team turned to multivariate logistic regression, a method that estimates the odds of belonging to a given profile while adjusting for other variables simultaneously. Six factors emerged as significantly associated with profile membership: age, education level, prior AI training experience, frequency of AI use at work, attitudes toward AI, and perceived organizational support. Nurses holding a bachelor’s degree or higher were more likely to occupy the stronger profiles, as were younger respondents, though the relationship with age is not simply a matter of digital nativity, since it interacts with career stage and exposure to continuing education.

Two of the associated factors stand out for their practical implications. The first is training experience: nurses who had participated in any structured AI training were far more likely to land in the higher-literacy profiles, which makes targeted training for those without such experience an obvious priority. The second is frequency of use, which points to a self-reinforcing loop. Nurses who encounter AI tools regularly build familiarity and confidence, which in turn encourages further use, while those working on units without deployed AI systems have few opportunities to practice and risk falling permanently behind. The study’s findings suggest that access itself, not merely motivation, may be a bottleneck for literacy development.

The psychological and organizational dimensions proved equally consequential. Using validated instruments to measure attitudes toward AI in the workplace and organizational support, the researchers found that nurses with more positive attitudes and stronger institutional backing were significantly more likely to belong to the high-literacy profiles. This aligns with a broader body of technology-acceptance research showing that perceived usefulness and perceived organizational backing are powerful predictors of adoption behavior. In practical terms, a hospital that communicates a clear AI strategy, provides time and resources for learning, and frames AI as augmentation rather than surveillance is likely to cultivate a more literate workforce than one that simply installs software and expects adaptation.

The authors flag two groups for special attention: nurses with lower educational attainment and those without prior AI training, who they note may encounter greater obstacles in developing AI literacy. This matters for equity as much as for efficiency. If AI competence accrues disproportionately to the already advantaged, hospitals risk creating a two-tier clinical workforce in which some nurses can leverage intelligent decision support while others are locked out of it, with consequences for career progression and potentially for patient safety. Deliberately designed bridging programs, delivered at the bedside and tailored to different starting points, could interrupt that divergence before it hardens.

The study carries the usual caveats of its design. As a cross-sectional survey conducted between January and April 2026, it captures a snapshot rather than a trajectory, so the causal direction of the associations cannot be established with certainty; it is possible, for instance, that positive attitudes drive AI use, or that successful use breeds positive attitudes, or both. The sample, while large and drawn from multiple centers, is confined to China’s hospital system, and the translation of AI literacy profiles to other health systems with different technology infrastructures and nursing education pathways remains an open question. Nevertheless, the methodological shift from averaging nurses to sorting them is the study’s real contribution, and it is one likely to be imitated. As AI penetrates triage, monitoring, documentation, and diagnosis, health systems worldwide will need to know not just how much AI literacy their nurses have, but which kinds, in whom, and under what conditions it grows. This study offers one of the first rigorous maps of that landscape, and its message is clear: the AI-ready nurse is not yet the norm, but the factors that produce one, training, practice, attitude, and institutional support, are all within reach of deliberate policy.

Subject of Research: Artificial intelligence literacy profiles and associated factors among clinical nurses in China

Article Title: Potential categories and associated factors of artificial intelligence literacy in Chinese nurses: a multicenter cross-sectional study

Article References: Potential categories and associated factors of artificial intelligence literacy in Chinese nurses: a multicenter cross-sectional study. (n.d.). https://doi.org/10.1186/s12912-026-05406-0

Image Credits: AI Generated

DOI: 10.1186/s12912-026-05406-0

Keywords: artificial intelligence literacy, nursing, latent profile analysis, cross-sectional study, digital health, clinical decision support, health workforce, organizational support, AI training, China, technology acceptance, nurse education