Does digital literacy prepare nursing students for medical AI?
A two-country survey linked stronger digital literacy with higher self-reported AI readiness in 500 nursing students. Its cross-sectional design cannot show that literacy training caused readiness—or that students can use clinical AI safely.
The short version
- What happened
- The researchers systematically sampled 510 students from about 10,200 eligible undergraduates at two universities; 500 completed the paper questionnaire.
- Who it affects
- Nursing students need practice in evidence checking, privacy, escalation and recognising unsafe outputs—not only confidence with digital tools.
- What remains unproven
- Cross-sectional design cannot establish that competence or literacy caused AI readiness
Keep in mind
Cross-sectional design cannot establish that competence or literacy caused AI readiness
All three main constructs were self-reported rather than observed in clinical tasks
Two-university sample limits national, regional and international generalisability
Students with specialised AI training were excluded, narrowing the range of prior experience
The AI-readiness scale was designed for medical students and was adapted because no fully validated nursing-specific measure was available
Source analysis · 1 linked source. Primary analysed source date: 15 October 2026.
Read the full report · 8 min The evidence, methods, implications and limitations
The answer: literacy tracked readiness, but the study cannot prove training works
Nursing students who reported stronger digital competence and digital literacy also reported greater readiness to engage with medical AI. In a peer-reviewed PLOS One study of 500 undergraduates in Saudi Arabia and Egypt, AI-readiness scores correlated with digital competence at r=0.749 and with digital literacy at r=0.663. A path model found that digital literacy partially mediated the association between competence and readiness, meaning the data were consistent with both a direct link and an indirect link through literacy.
The result does not show that a digital-literacy course will make a student clinically safe with AI. The survey captured self-assessments at one time point, not observed practice before and after an intervention. Students did not have to identify a hallucinated citation, challenge an unsafe recommendation, protect patient data or decide when to escalate to a clinician. The practical message for educators is therefore to test a promising curriculum target, not to treat confidence or readiness scores as evidence of competence.[1]
Who took part and how the sample was selected
The study covered two universities for health sciences, one in Saudi Arabia and one in Egypt. Eligible students were enrolled in any undergraduate nursing year, had completed at least one semester, had basic access to a phone, tablet or computer, and could complete the survey in Arabic or English. Students on academic leave, in bridging or external programmes, or with previous specialised AI-only training were excluded. Excluding AI-trained students reduced one source of prior exposure but also removed some of the people most likely to demonstrate advanced readiness.
The researchers estimated a minimum sample of 459 using an anticipated effect size of 0.15, alpha of 0.05 and power of 0.90. About 10,200 students were eligible across the two institutions. The team used a predetermined list and selected every twentieth student, producing an initial sample of 510; ten declined, leaving 500 participants. Data were collected on paper between 29 October and 24 December 2025. The article reports ethics approvals at both universities and written informed consent.[1]
Three self-report scales defined the result
Digital competence was measured with a 28-item scale covering information search, content creation, communication and collaboration, storage and management, evaluation and protection. Digital literacy used a 20-item scale about accessing, analysing, producing and communicating digital information. Medical-AI readiness was measured with the 22-item MAIRS-MS instrument, which covers cognition, ability, vision and ethics. That instrument was developed for medical students; the authors used it because no fully validated nursing-specific readiness scale was available.
Bilingual specialists translated the instruments into Arabic and back-translated them, experts reviewed face validity, and a 50-student pilot tested clarity and reliability; those students were excluded from the final sample. The competence and AI-readiness measures each produced a Cronbach alpha of 0.91 in this study, indicating high internal consistency. Reliability does not establish that a scale measures safe clinical use, however. Closely related self-report constructs can correlate strongly because they share language, response styles and perceptions of confidence.[1]
What the correlations and mediation model found
Digital competence correlated with digital literacy at r=0.724. AI readiness correlated with digital competence at r=0.749 and digital literacy at r=0.663. Within the readiness scale, the cognition, ability and vision domains correlated strongly with the total readiness score, as expected for components of the same measure. All reported correlations in that table were statistically significant at the 0.01 level.
In the path analysis, digital competence was associated with digital literacy (standardised beta 0.813). When both were included as predictors, competence had a beta of 0.360 and literacy a beta of 0.457 for AI readiness. The reported indirect effect through literacy was beta 0.371, with a bootstrapped 95% confidence interval from 0.268 to 0.475. The authors describe this as partial mediation. Because exposure, mediator and outcome were measured together, the temporal order required for causal mediation is not demonstrated; readiness could also shape how students rate their skills.[1]
Exposure and interest were associated with higher scores
Students with prior exposure to digital tools reported higher scores in all three domains. Those who said they always used digital tools had the highest digital-competence and AI-readiness scores, and students reporting very high interest in emerging technologies scored 113.5 on competence and 88.8 on readiness. Students older than 20 also had higher average competence and readiness scores, while male students reported slightly higher digital literacy. Employment and parental education were associated with some measures.
These subgroup comparisons are descriptive associations, not proof that device use, employment or enthusiasm caused readiness. Interest can influence both experience and survey responses, while opportunities to use technology may reflect socioeconomic or educational advantage. The model reported no statistically significant covariate associations with AI readiness after adjustment, although interest in emerging technologies approached the authors’ threshold. Educators should avoid turning these results into demographic stereotypes or assuming frequent device use equals critical digital judgement.[1]
What nursing educators can do now
The defensible next step is to integrate digital literacy with authentic clinical reasoning rather than add a generic AI-awareness lecture. Students can practise checking whether a cited paper exists, matching its population and outcome to the patient, identifying missing clinical context, documenting consequential use and recognising when privacy rules prohibit entering data. Assessment should include deliberately flawed outputs and require an explanation of why the answer should be accepted, modified, rejected or escalated.
Curricula should also separate four capabilities that a single readiness score can blur: conceptual understanding of AI, practical operation of a tool, critical appraisal of evidence and ethical action under clinical accountability. Nursing students need opportunities to work with informatics specialists, patients and practising nurses, but the goal is not tool enthusiasm. It is reliable performance when an output conflicts with observation, a patient preference or local protocol. Faculty development and access to approved systems are prerequisites; placing the burden on students alone would be unsafe.[1]
What health services should not infer
Hospitals and regulators should not use this survey to certify workforce readiness or justify autonomous AI use. No participant was observed in a clinical task, no patient information was processed and no error, escalation or outcome rate was measured. Higher self-reported readiness may support learning, but it can also coexist with overconfidence. A safe transition from university to practice requires supervised use, clear scopes of responsibility, access to source evidence and monitoring that distinguishes a useful challenge from silent acceptance.
The two-country design is valuable because much AI-education evidence is concentrated in Europe and North America. Even so, two universities cannot represent Saudi Arabia, Egypt, the Middle East and North Africa, much less global nursing education. Programme structure, language, digital infrastructure and clinical placement opportunities may differ substantially. The paper supports local curriculum research and cross-country replication, not a universal claim about nurses.[1]
Limits, funding and what would change the assessment
The authors identify the cross-sectional design, self-report bias, two-university sample and limited regional generalisability as key limitations. Excluding students with specialised AI training may have narrowed the range of readiness. The study also adapted a medical-student AI-readiness scale because a validated nursing-specific alternative was unavailable. The paper reports no specific funding and no competing interests, and says the underlying data are available in the article and supporting information.
Confidence would rise with a preregistered trial in which nursing programmes are assigned to a defined curriculum and students complete validated tasks before training, immediately afterwards and during clinical placements. Outcomes should include evidence verification, privacy decisions, detection of unsafe recommendations, calibrated confidence, escalation behaviour and performance across languages and access needs. Follow-up with patients and supervisors could test whether skills transfer without increasing workload or inequity. Until such evidence exists, digital literacy is a plausible and measurable foundation—not proof of safe medical-AI practice.[1]