GENETIC COUNSELING WITH AI CHATBOTS, RISK BENEFIT ANALYSIS

Objective: To assess the potential benefits and risks of artificial intelligence (AI) applications in genetic counseling by evaluating the performance of commonly used AI platforms on simulated patient scenarios.Materials and Methods: Nine simulated patient datasets were created to reflect typical patient profiles encountered in routine genetic counseling. Responses generated by the AI platforms were evaluated according to current guidelines related to the relevant diseases. Each response was classified as sufficient, insufficient, irrelevant, or misinformation.Results: Of 218 total responses, 152 (69.72%) were deemed sufficient, 34 (15.60%) insufficient, 13 (5.96%) irrelevant, and 19 (8.72%) misinformation. Among the sufficient responses, ChatGPT-4 showed the highest rate at 78.57%, followed by ChatGPT-3.5 at 76.79%, Gemini at 71.43%, and Copilot at 50.00%.Conclusion: AI platforms developed using reliable data and designed with transparency and observability can improve access to genetic counseling and support individual informational needs. However, important risks remain, including data misinterpretation, inaccurate responses, misinformation, noise in training data, hallucinations, and privacy concerns. Overall, while these AI platforms show promise in providing foundational clinical information, their frequent omissions, unconfirmed assumptions, and critical technical or clinical inaccuracies underscore significant limitations in their current utility for reliable genetic counseling.