Bridging the Gap: Evaluating Publicly Available Education on SSRI Use in Pregnancy in the Age of Artificial Intelligence [ID 3729]

INTRODUCTION: In the context of declining literacy rates and July 2025 FDA scrutiny of selective serotonin reuptake inhibitor (SSRI) use in pregnancy, this study compares the readability of relevant patient education across medical societies, educational websites, and social media, and evaluates large language models (LLMs) as tools for improving patient understanding. METHODS: Patient-facing content on SSRI use in pregnancy from the American Academy of Family Physicians, the American College of Nurse-Midwives, the American College of Obstetricians & Gynecologists, and the Society for Maternal-Fetal Medicine was compared to educational websites, Reddit, Facebook, and LLM-generated responses (ChatGPT 4.0 and Gemini 2.5). Readability was assessed using five indices (Flesch–Kincaid, Gunning Fog, ARI, Coleman–Liau, and SMOG), and an average grade level was calculated. Large language models were asked to rewrite ACOG’s FAQs on anxiety and depression in pregnancy, and their outputs were analyzed. RESULTS: Reddit posts were the most readable (grade 9.58, SD 1.59), followed by Facebook (9.97, SD 1.63) and educational websites (13.11, SD 1.13). Among medical societies, the ACOG statement was at college reading level (15.76, SD 1.89), lower than SMFM’s (18.49, SD 1.38), AAFP’s (17; SD 1.25), and ACN-M’s (17.50, SD 1.33), which were all college graduate level. Responses from Gemini (13.99, SD 1.37) and Chat GPT (15.3, SD 1.37) were more readable than ACOG FAQs (10.79, SD 1.72). The ACOG FAQs rewritten by LLMs were the most readable (Chat GPT 8.95, SD 1.18; Gemini 8.78, SD 1.69), although still not at the requested sixth-grade level. CONCLUSIONS/IMPLICATIONS: More readable patient education on SSRI use in pregnancy is needed, as most Americans read at or below a seventh-grade level. Large language models may be useful in addressing this gap.