ChatGPT told me I was pregnant. It was dead wrong.

Published: Friday, July 24, 2026 at 11:00 am

AI Chatbots Offer Confident, Yet Potentially Dangerous, Medical Advice

Recent experiences with artificial intelligence chatbots, including the launch of ChatGPT Health, underscore a critical concern: the unreliability of AI for medical diagnoses. Users are reporting instances where AI has provided definitive, yet incorrect, health advice, highlighting the inherent risks of relying on these tools for medical guidance.

One user described inputting a series of symptoms, including irregular cycles, night sweats, mood swings, and exhaustion, into ChatGPT. The AI confidently diagnosed pregnancy, a conclusion that was factually impossible for the user. This misdiagnosis, while not immediately life-threatening, illustrates the AI's tendency to prioritize statistical probability over individual context. The AI's algorithms, trained on vast amounts of text data, identified pregnancy as a common association with the presented symptoms, failing to account for factors like perimenopause or a history of vasectomy. The user emphasized that such confident, yet erroneous, advice could have significant emotional and practical consequences for individuals facing health uncertainties.

This issue is not isolated. A lawsuit filed against OpenAI, the creator of ChatGPT, by a Florida pastor, Scott Winters, details a near-fatal experience. After consulting the chatbot about severe dizziness and pelvic pain, Mr. Winters was assured his symptoms were not dangerous and advised to rest. Weeks later, he was hospitalized with a life-threatening pulmonary embolism, a condition his doctors stated was worsened by the prolonged immobility recommended by the AI.

Experts explain that large language models are designed to generate text that sounds authoritative, not necessarily to be medically accurate. AI operates by pattern-matching words and predicting the most statistically probable next word based on its training data. It lacks the capacity for genuine medical understanding, clinical judgment, or the ability to consider nuanced physiological realities. Furthermore, AI exhibits a tendency towards "sycophancy," mirroring a user's language and assumptions, which can lead to confirmation bias and reinforce inaccurate beliefs. The authoritative tone adopted by AI, even when delivering incorrect information, can create a false sense of security, potentially leading users to disregard professional medical advice or ignore serious warning signs.

While AI can be a useful tool for organizing thoughts or summarizing medical information, it is not a substitute for human medical professionals. The consensus is that AI lacks the ability to perform physical examinations, order tests, or provide the nuanced care that a licensed doctor can. For anyone experiencing concerning symptoms, the recommendation is to consult a human healthcare provider rather than relying on AI for diagnoses.

BNN's Perspective: The rapid advancement of AI technology presents both opportunities and significant challenges. While the potential for AI to assist in various fields is undeniable, its application in healthcare demands extreme caution. The instances highlighted underscore the critical need for users to maintain a healthy skepticism when interacting with AI for medical advice. The authoritative tone of AI can be misleading, and the underlying mechanisms of language models are not designed for the complex and individualized nature of human health. Prioritizing consultation with qualified medical professionals remains the most responsible and safest approach to managing one's health.

Tags: ChatGPT, AI, medical advice, diagnosis, pregnancy, perimenopause, pulmonary embolism, lawsuit, OpenAI, large language models, symptoms, healthcare, reliability, accuracy, technology

Full Story