EVALUATING THE ACCURACY, COMPLETENESS, AND READABILITY OF ARTIFICIAL INTELLIGENCE–GENERATED INFORMATION ON COMMONLY PROMOTED FERTILITY SUPPLEMENTS IN ENGLISH AND SPANISH

Objective: To evaluate the accuracy, completeness, safety counseling, and readability of artificial intelligence (AI)–generated patient-facing information on commonly used fertility supplements in English and Spanish. Design: Cross-sectional content analysis. Materials and Methods: ChatGPT version 5.5 was queried regarding 14 commonly promoted fertility supplements identified from recurring recommendations across consumer-facing fertility resources, including online marketplaces and fertility websites: coenzyme Q10 (CoQ10), myo- inositol, dehydroepiandrosterone (DHEA), folic acid, omega-3 fatty acids, N-acetylcysteine (NAC), melatonin, zinc, maca, ashwagandha, alpha-lipoic acid, D-aspartic acid, choline, and iodine. Five standardized prompts were submitted in English and Spanish addressing use, efficacy, risks, pregnancy safety, and patient counseling. Responses were compared with an evidence-based reference. Three bilingual reviewers independently scored responses on a 5-point Likert scale across four domains: accuracy, completeness, safety counseling, and readability. Inter-rater agreement was assessed by agreement within one point. Results: A total of 140 AI-generated responses were evaluated. English responses scored higher than Spanish responses across all evaluated domains, including accuracy (3.79 vs. 3.64), completeness (3.43 vs. 3.21), and readability (3.93 vs. 3.58). Safety counseling remained limited in both languages and was lowest in Spanish (3.31). Folic acid and myo-inositol had the highest overall scores in both languages, while Spanish-language D-aspartic acid responses had the lowest completeness score (2.67). Inter-rater reliability showed close reviewer agreement, with 100% agreement within one point across all domains; ICC analysis demonstrated fair exact-score reliability, likely due to limited score variability. Conclusions: AI–generated fertility supplement information is generally readable but demonstrates meaningful gaps in completeness and safety counseling, particularly in Spanish. These findings suggest AI tools may reinforce language-based disparities in reproductive health information. Clinicians should counsel patients that AI-generated information does not replace individualized medical guidance, especially when supplements may interact with fertility treatment, pregnancy, endocrine disorders, or hormone-sensitive conditions.