Nursing students felt more AI-aware than AI-capable in a two-university survey
Among 684 nursing students in Hubei, self-rated AI cognition exceeded competency and only 25.6% had taken AI coursework. The survey did not test clinical or technical skill.
The short version
- What happened
- Researchers received 684 valid questionnaires from 723 nursing students at two Hubei universities; 574 respondents were women and the mean age was 19.44.
- Who it affects
- Nursing students need supervised opportunities to practise verification, privacy and escalation—not just access to AI tools—before readiness claims are used to shape clinical responsibilities.
- What remains unproven
- Cross-sectional design cannot establish causal relationships
Keep in mind
Cross-sectional design cannot establish causal relationships
Convenience sample from two Hubei universities
Entirely self-reported measures
Scale construct validity not established specifically for nursing students
Unmeasured school, faculty and family technology factors
No observed clinical skill, safety or patient outcome
Source analysis · 1 linked source. Primary analysed source date: 26 October 2026.
Read the full report · 3 min The evidence, methods, implications and limitations
Awareness outpaced confidence in using AI
A peer-reviewed survey of 684 nursing students at two universities in Hubei, China, found a gap between understanding AI and feeling capable of applying it. On the five-point Chinese Medical Artificial Intelligence Readiness Scale, the overall mean was 3.47. Cognition scored highest at 3.57, ethics 3.52, vision 3.45 and competency lowest at 3.32.
Only 175 students—25.6%—said they had participated in AI coursework, while 327, or 47.8%, reported using large language models frequently or regularly for learning. Those figures give educators a concrete reason to examine whether informal tool use is running ahead of structured instruction. They do not show that the students could choose, verify or safely use AI in clinical care.[1]
Associations point to training questions, not causes
The researchers distributed 723 questionnaires in April and May 2026 and retained 684, a 94.6% valid-response rate. Their regression explained 31.1% of variance in readiness scores. More frequent LLM use, higher education level, AI-course participation and programming foundations were among the associated factors; only 107 students, 15.6%, reported a programming foundation.
The sample was recruited by convenience from two universities, all measures were self-reported and the scale was developed for medical students; its construct validity has not been established specifically for nursing students. Confidence would rise with multi-centre sampling, observed AI-use tasks, clinical scenarios and longitudinal measures. The study received a university teaching grant, declared no commercial or financial conflicts, and disclosed using DeepSeek, Kimi and GPT-4 only for English translation and language polishing.[1]
The distinction that matters
Readiness is not the same as clinical competence
The survey describes students’ own ratings across four domains; it did not test performance with patients or clinical records.
Cognition
Self-rated understanding
Highest mean: 3.57 of 5
Competency
Self-rated ability to apply AI
Lowest mean: 3.32 of 5
Clinical practice
No observed task or patient outcome
Safety and effectiveness remain unknown