Clinician Acceptance of Artificial Intelligence- and Extended Reality-Enabled Telemedicine: A Cross-Sectional Vignette Survey of Residents and Nurses in Romania.

Background and Objectives: Artificial intelligence (AI) and extended reality (XR) are increasingly being integrated into telemedicine, yet clinician adoption depends not only on perceived utility but also on digital preparedness and technology-related burden. This study compared clinician acceptance of AI-only, XR-only, and combined AI+XR telemedicine scenarios and examined whether AI literacy, health literacy, technostress, age, and sex explained variability in acceptance. Materials and Methods: We conducted a cross-sectional, anonymous online vignette survey among 117 resident physicians and nurses from a tertiary-care hospital and affiliated outpatient clinics in Western Romania. Participants evaluated three randomized telemedicine scenarios (AI-only, XR-only, and AI+XR) using a 3-item Acceptance Index scored from 1 to 4. Additional measures included a study-developed AI literacy quiz, the Romanian-validated HLS-EU-Q16, an adapted brief technostress scale, prior AI/XR exposure, perceived accessibility, perceived value, privacy concern, and demographic variables. Results: Acceptance was highest for the combined AI+XR scenario (3.74 ± 0.49), followed by AI-only (3.64 ± 0.56) and XR-only (3.49 ± 0.64). AI+XR acceptance was significantly higher than AI-only and XR-only acceptance (both p < 0.001), although the absolute between-scenario differences were modest. Residents reported consistently higher acceptance than nurses across all scenarios, whereas sex differences were small and non-significant; younger age showed only weak inverse associations with acceptance. AI+XR acceptance correlated positively with AI literacy (ρ = 0.60) and health literacy (ρ = 0.23), and negatively with technostress (ρ = -0.47). In multivariable analysis, higher AI literacy (β = 0.603, p < 0.001) and health literacy (β = 0.241, p < 0.001) independently predicted higher AI+XR acceptance, whereas technostress was inversely associated (β = -0.212, p < 0.001). Conclusions: In this sample, clinicians rated the integrated AI+XR vignette most favorably, but the observed differences between scenarios should be interpreted cautiously and as hypothesis-generating rather than definitive evidence of superiority. Acceptance appeared to depend more on digital readiness and technostress than on age or sex, supporting implementation strategies focused on literacy-building, workflow fit, and burden reduction.