A nationwide cross-sectional study of 1,136 practicing nurses across the United States revealed that workplace supervisory and organizational support were positively associated with generative artificial intelligence use, while having a workplace policy showed no significant association. The survey findings indicate that perceived supervisory and organizational support are positively associated with nurses' reported use of generative artificial intelligence tools.
Artificial intelligence technologies have steadily penetrated modern clinical environments, expanding from experimental pilot programs into foundational hospital infrastructure. Prior health services evaluations, such as analyses of the 2023 American Hospital Association Annual Survey, noted that approximately 65 percent of hospitals in the United States used predictive models to predict inpatient health trajectories, identify high-risk outpatients, monitor health, and automate scheduling and billing. Among adopting hospitals, approximately 79 percent relied on models supplied by their electronic health record developer. In sharp contrast, generative artificial intelligence platforms such as ChatGPT, Claude, and Gemini present a fundamentally distinct implementation paradigm. Because general-purpose tools such as ChatGPT, Claude, and Gemini are freely available to individual clinicians, they can be applied to clinical questions, patient education, and documentation tasks largely outside institutional oversight.
When adoption outpaces governance, these tools can introduce quantifiable failure modes, and clinicians have raised concerns regarding data privacy, legal liability, algorithmic bias, and the erosion of human oversight. Controlled evaluations of machine-generated medical literature have demonstrated that up to 47 percent of generated references were entirely fabricated, while another 46 percent were authentic citations applied inaccurately, leaving only a meager 7 percent both authentic and accurate. Additional investigations testing commercially available models across realistic clinical vignettes revealed that algorithms frequently perpetuated debunked, race-based medical misconceptions and displayed erratic variability when identical clinical prompts were queried repeatedly.
Despite mounting technical scrutiny, empirical research regarding how practicing nurses actually navigate generative artificial intelligence within day-to-day employment has lagged considerably behind broader technological integration.
To evaluate these dynamics, the researchers recruited a diverse national nonprobability sample of practicing, licensed nurses utilizing the Centiment research panel between June 5 and July 3, 2026. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines and received institutional review board approval. From an initial export of 1,239 completed submissions, rigorous quality-control filters excluded respondents who lacked active nursing licensure, were not actively employed in healthcare roles, did not provide or supervise direct clinical care, failed embedded attention-check questions, or declined informed consent.
Descriptive profiling indicated that 81.5 percent of participants identified as female, with 50.3 percent holding a baccalaureate nursing degree and an additional 19.1 percent possessing graduate credentials such as a master’s degree, Doctor of Nursing Practice, or Doctor of Philosophy. Bedside nurses constituted 33.9 percent of the cohort, accompanied by charge nurses at 23.4 percent, advanced practice registered nurses at 19.3 percent, and nurse managers or administrators at 15.9 percent. Approximately 63.4 percent practiced in acute care hospital settings, with clinical experience spanning from novice practitioners to seasoned veterans possessing more than twenty years of bedside experience.
The study measured two distinct behavioral outcomes: any work-related generative artificial intelligence usage within the prior 12 months, and the habitual frequency of usage specifically dedicated to clinical patient-care tasks. Analysis revealed that 65.3 percent of respondents, representing 742 clinicians, had utilized generative tools for occupational duties over the preceding year. When evaluating clinical intensity on an unbounded habitual frequency scale, 73.9 percent acknowledged at least occasional clinical application, with 46.5 percent using generative models weekly or more, and 23.3 percent reporting daily or near-daily usage. Cross-tabulations confirmed that these two metrics operated under distinct cognitive recall frames rather than strictly nested hierarchical subsets, leading researchers to treat the frequency metric as a dedicated measure of clinical intensity rather than an exact annualized prevalence benchmark.
Multivariable binary logistic regression revealed that frontline leadership and institutional backing were among the most potent correlates of work-related adoption. After adjusting for demographic factors, professional roles, and clinical settings, every single-point increase on a five-point Likert scale of supervisory support was associated with a 42 percent increase in the adjusted odds of utilizing generative tools. Similarly, each additional point of perceived organizational support corresponded to a 43 percent increase in adjusted odds. Because the study design was cross-sectional, these associations do not establish causality, as nurses who use generative AI may perceive support differently from non-users.
The investigation also captured an intricate relationship between workplace policy communication and clinician behavior. Approximately 47.0 percent of respondents reported that their employer possessed an active policy governing artificial intelligence in patient care, whereas 32.5 percent reported no policy, and 20.5 percent were uncertain whether guidelines existed. In multivariable models, the mere confirmed existence of a policy was not significantly associated with reduced or increased adoption compared to having no policy at all. However, nurses who did not know whether an institutional policy existed demonstrated significantly lower odds of both general work-related use and frequent clinical deployment. This pattern suggests that policy communication, not merely policy existence, warrants attention in future implementation research.
A parallel gradient characterized formal training status, which researchers evaluated as a composite construct combining completed training with personal interest among untrained personnel. Half of the surveyed nurses, exactly 50.4 percent, had completed formal educational modules on artificial intelligence. Among the remaining participants, 31.6 percent expressed an explicit desire for training, while 18.0 percent remained untrained and uninterested. In multivariable models, untrained clinicians lacking interest demonstrated dramatically lower odds of generative artificial intelligence usage compared to trained peers. The authors emphasize that the training-status variable reflects training receipt and interest among untrained respondents, rather than a causal estimate of training effectiveness.
Professional clinical roles also strongly patterned technological uptake across healthcare facilities. In contrast, standard demographic variables including age, gender, years of nursing experience, formal educational attainment, and geographic census regions displayed no statistically significant independent associations with adoption once workplace governance and clinical roles were accounted for.
The authors emphasize several methodological limitations inherent to cross-sectional survey designs that necessitate cautious interpretation of the empirical conclusions. Importantly, the survey assessed self-reported frequency rather than verified output quality, patient outcomes, adherence to data privacy regulations, or clinical safety. The researchers conclude that future longitudinal and prospective studies are needed to establish the direction of these associations, test whether policy communication or training changes behavior, and directly measure verification behaviors and safety outcomes.
Subject of Research: Generative artificial intelligence adoption patterns and workplace governance correlates among practicing United States nurses
Article Title: Workplace Policy Awareness, Training, and Supervisory Support as Correlates of Generative Artificial Intelligence Use Among Practicing Nurses
Article References: Carter, G. A. (2026). Workplace Policy Awareness, Training, and Supervisory Support as Correlates of Generative Artificial Intelligence Use Among Practicing Nurses. Nursing Open, 13(11), Article e71000. https://doi.org/10.1002/nop2.71000
Image Credits: AI Generated
DOI: 10.1002/nop2.71000
Keywords: generative artificial intelligence, nursing workforce, workplace policy, supervisory support, organizational support, clinical documentation, health informatics, patient safety, cross-sectional survey, clinical decision support, artificial intelligence training, hospital nursing, advanced practice nursing, large language models, implementation science, nursing administration
News Source: Ophelia Keating. (November 14, 2026). Workplace Support Correlates With Generative AI Use Among US Nurses. Scienmag.
