Accuracy and Readability of Generative Artificial Intelligence for Vascular Surgery Patients: A Specialist Based Evaluation Highlighting the Current Landscape of Safety Risks and Accessibility Gaps.

No significant differences were found in tone or complementarity across disease categories or question types.

Urgency did not differ significantly between symptom related and treatment related queries overall; however, urgency varied significantly across subtypes of symptom related questions (p = .03), consistent with inconsistent escalation recommendations.

The mean Flesch Reading Ease was 32.3 ± 12.1 and Flesch-Kincaid Grade Level 13.5 ± 2, corresponding to a university level reading requirement.

Treatment related answers were particularly weak, with only 25% deemed appropriate.