Age-Stratified Evaluation of ChatGPT Recommendations for Abnormal Uterine Bleeding: An Expert-Based Assessment of Guideline Adherence and Clinical Appropriateness

Background: Abnormal uterine bleeding (AUB) requires age-sensitive evaluation because endometrial hyperplasia or malignancy risk increases with age and after menopause. Large language models are increasingly used for health information, but their performance in age-dependent gynecologic triage remains unclear.Objective: To evaluate the clinical appropriateness, guideline adherence, safety, completeness, endometrial sampling accuracy, malignancy-risk recognition, and urgency classification of ChatGPT responses to standardized AUB and postmenopausal bleeding scenarios.Methods: Sixty investigator-generated scenarios were reviewed by two gynecologic oncologists and stratified into five groups: adolescents, reproductive-age women, late reproductive-age women, premenopausal/perimenopausal women aged 45–55 years, and women with postmenopausal bleeding. Each scenario was submitted to ChatGPT using a standardized prompt. Responses were independently evaluated by two gynecologic oncologists across four domains: clinical appropriateness, guideline adherence, patient safety, and completeness.Results: The overall expert-rated appropriateness score was 3.48 ± 0.66, and 21 of 60 responses (35.0%) were appropriate. Mean overall scores decreased from 4.10 ± 0.41 in adolescents and 3.92 ± 0.33 in reproductive-age women to 2.98 ± 0.49 in women aged 45–55 years and 2.82 ± 0.51 in postmenopausal bleeding scenarios (p < 0.001). Endometrial sampling recommendations were correct in 72 of 120 reviewer-level assessments (60.0%). High-risk or malignancy-related features were recognized in 19 of 40 relevant assessments (47.5%). Inter-rater reliability for overall score was excellent (ICC = 0.957).Conclusion: ChatGPT performed better in adolescent and reproductive-age AUB scenarios but showed lower appropriateness in age groups with higher endometrial pathology risk. Clinician oversight remains necessary when AI-generated recommendations may influence AUB triage or endometrial cancer risk assessment.