Workplace Policy Awareness, Training, and Supervisory Support as Correlates of Generative Artificial Intelligence Use Among Practicing Nurses

To estimate the prevalence and frequency of generative artificial intelligence (AI) use among practicing nurses and to determine whether four workplace governance conditions were independently associated with use.

US-wide nonprobability cross-sectional survey.

Licensed, practicing nurses in the United States (N = 1136) recruited through a commercial, US-wide nonprobability online panel completed a self-administered questionnaire. Any work-related use in the previous 12 months was modelled using binary logistic regression, and the frequency of clinical use, measured with an unbounded habitual-use item, was modelled using proportional-odds ordinal regression. Both models included workplace policy awareness, training status (training receipt and interest among untrained respondents), supervisor support, and organizational support and adjusted for role, setting, education, experience, age, gender, and geographic region. Sensitivity analyses examined ordinal-covariate coding, the proportional-odds assumption, and support-item specification.

In the previous 12 months, 742 nurses (65.3%) reported using a generative AI tool for a work-related task, and 528 (46.5%) reported at least weekly use for clinical tasks on a separate habitual-frequency item without a bounded recall period. The two estimates are not nested because the frequency item used a different reference frame. Each point of supervisor support was associated with higher odds of any use (aOR = 1.42, 95% CI [1.17, 1.72]) and of more frequent clinical use (aOR = 2.05, 95% CI [1.75, 2.40]), with parallel associations for organizational support (aOR = 1.43 and 1.60). Nurses who were untrained and uninterested in training had markedly lower odds of both behaviours than trained nurses (aOR = 0.16 and 0.24); this comparison combines training receipt and interest and is not an estimate of training effectiveness. Uncertainty about whether a workplace policy existed, rather than its absence, was associated with lower use (aOR = 0.62 and 0.65). No statistically significant independent associations were detected for education, experience, age, gender, setting, or region, although role was associated with use.

In this US-wide nonprobability sample, generative AI use was common and associated with visible workplace governance conditions and professional role. The cross-sectional design supports correlational, not causal, conclusions.

Communicated policy, accessible training, and frontline supervisory support warrant evaluation as potential governance approaches; this study does not establish whether frequent use was safe, appropriate, or outcome-improving.

The study addressed the lack of national evidence linking workplace governance to nurses' reported use of generative AI. Use was common and associated with supervisor support, training status (receipt and interest among untrained respondents), policy awareness, and professional role. The findings will inform nurses, nurse managers, health system leaders, and policy bodies responsible for evaluating and governing these tools.

Reporting followed the STROBE guidelines for cross-sectional studies.

No Patient or Public Contribution. This survey of practicing nurses did not involve patients or members of the public in its design, conduct, analysis, or the preparation of the manuscript.