Abstract
BACKGROUND: Artificial intelligence (AI) could reshape healthcare delivery, but its adoption depends on nurses' attitudes, literacy, readiness, and intentions. METHODS: Following PRISMA 2020, we searched six databases from inception to May 2025 and undertook thematic synthesis. A non-systematic horizon scan (June-August 2025) informed interpretation only. RESULTS: Thirty-seven studies met inclusion: 28 analytical cross-sectional surveys, 8 qualitative studies, and 1 quasi-experimental trial.Nursing students generally held moderately positive attitudes towards AI; senior students were more enthusiastic than juniors, and men more than women. Students reported moderate literacy and readiness; prior AI training and stronger computer skills correlated with more favourable attitudes and greater adoption intentions, whereas anxiety dampened readiness. Many students doubted AI's ability to outperform humans in routine tasks and flagged integrity risks, underscoring the need for age-appropriate instruction and safeguards. Practising nurses expressed moderate safety and error concerns but showed greater optimism among younger staff; across studies, nurses consistently argued AI should augment-not replace-human empathy and judgement. Targeted training substantially improved, and largely maintained, AI knowledge; leadership endorsement and phased, user-centred roll-outs strengthened readiness, while outdated infrastructure, resource constraints, ethical/privacy concerns, and fear of deskilling impeded progress. Determinants of attitudes and intentions clustered around perceived usefulness/performance and effort expectancy, self-efficacy, digital literacy, and facilitating conditions. The horizon scan added signals of a preparedness-impact gap among nurse leaders, syllabus/policy language as a faculty readiness multiplier, role-specific adoption gaps (e.g., lower use among head nurses despite positive attitudes), and coexistence of high AI anxiety with positive attitudes in students. CONCLUSION: Global nursing exhibits guarded optimism grounded in moderate literacy and readiness yet constrained by infrastructural, ethical, and pedagogical barriers. Adoption is driven by perceived usefulness, self-efficacy, and enabling environments, with anxiety and demographics moderating engagement. Priorities include embedding longitudinal AI competencies in curricula, iterative hands-on training, robust governance/ethics, and modernised infrastructure. Evidence dominated by cross-sectional designs and a narrow set of countries should be strengthened through longitudinal and experimental studies that validate psychometrics cross-culturally and link self-reports to objective use and patient-safety outcomes.