Health literacy and attitudes toward clinical AI: a cross-sectional study in Istanbul

To examine whether health literacy is associated with attitudes toward clinical artificial intelligence (AI) and to describe subgroup differences in this association.

We used two samples. A validation sample (N = 293) was used to develop and test the psychometrics of the Artificial Intelligence Knowledge, Usage, and Attitudes Scale (AI-KUAS). A separate application sample (N = 354) completed the 14-item Turkish Health Literacy Scale (HLS-14-TR) and the AI-KUAS. Analyses included correlations, group comparisons (t-tests/ANOVA with appropriate post-hoc procedures), and multiple linear regression adjusting for age, gender, and education. Two-tailed α = .05.

In the validation sample, a four-factor, second-order structure of the AI-KUAS was supported (CFI = .955, TLI = .943, SRMR = .037, RMSEA = .092), with high internal consistency (α = .961); a three-factor alternative was inferior (ΔCFI = .015; higher RMSEA). In the application sample, health literacy correlated moderately with AI attitudes (r = .51, p < .001) and remained a significant independent predictor in adjusted models (β = .39-.50, p < .001). Education showed additional positive associations with several subscales, whereas age and gender were not significant predictors. Model R² values ranged from .17 to .31 across outcomes.

Health literacy was positively and independently associated with more favorable attitudes toward clinical AI. Psychometric results suggested that perceived benefit and intention to use were closely related yet distinct dimensions within a four-factor structure. The findings align with literacy-aligned explanations, clear disclosure of AI use, and visible human oversight as practical implementation levers. Given the cross-sectional design and self-reported outcomes, longitudinal studies and measurement-invariance testing are warranted to support stronger causal and cross-group inference in future work.