P-387 Exploring the Potential of Artificial Intelligence in Providing Infertility Consultations and Answers: can it replace the infertility specialist? A hybrid study.

Abstract Study question What are the differences between accuracy of answers given by an AI-based infertility consultation system compared to those specialist? Summary answer OpenAI, AI platform, does not have potential provide proper women searching for regarding their infertility. is known already Artificial intelligence (AI) has been recently introduced in field human reproduction, both controlled ovarian stimulation algorithms and laboratory decisions. Additionally, AI-enabled technology used facilitate online communication patients healthcare professionals. However, recent widespread use chatbots causing serious concerns about providing medical advice. Furthermore, capacity intricate questions its treatment yet be explored. Despite being declared able accurate advice, a study assessing this conducted. design, size, duration The aim prospective observational was assess ability artificial appropriate asking questions. OpenAI generated provided specialists (control) were judged external, blinded terms scientific appropriateness (Correctness), detailed feature (Accuracy), properly (Precision) kindness (Empathy). Participants/materials, setting, methods 20 retrieved from web. For each question, openAI playground questioned “Answer question. You specialist”. platform available on internet website https://beta.openai.com. evaluators four different with more than years' experience they purpose presence answers. Four criteria (1-5): Correctness, Accuracy, Precision Empathy. Main results role chance mean total score without empathy 11.18/15 (14.53/20 empathy). 8.95/15 12.50/20 respectively. At Mann-Whitney test, scores showed statistically significant difference (p < 0.001). With tolerance -/+ 0.5 empathy, evaluated similar 4/20 answers, better specialist 2/20. Of outmost importance, none suspected engine tool. Intriguingly, evaluating absolute impact, obtained higher we expected great results. This result strongly suggests that, although overcome consultation, shown abilities mimicking specialist. usage or should considered due lack appropriateness, answering details, question-points precision. borderline performance human’s even at evaluation team specialists. Indeed, tool appropriately addressed worldwide spread. Limitations, reasons caution Further research needed understand limitations validity Caution ethical taken into consideration coming future. Wider implications findings also contributed develop study, giving our design. Moreover, intriguingly, abstract first hybrid-written (human AI) history ESHRE presentation. Trial registration number Not Applicable