AI Health Chatbots May Spread Dangerous Cancer Misinformation, Study Warns

A new study published in BMJ Open warns that artificial intelligence (AI) chatbots may provide misleading or potentially harmful medical advice, especially on sensitive topics such as cancer treatment. Researchers from Lundquist Institute for Biomedical Innovation at Harbor-UCLA Medical Center tested popular AI tools including OpenAI’s ChatGPT, Google Gemini, Meta AI, DeepSeek and Grok. Nearly half of the chatbot responses were found to be problematic, with some lacking proper context and others containing inaccurate or misleading information.

The study focused on common health questions people search online, including whether natural remedies can replace chemotherapy or whether products like 5G and antiperspirants cause cancer. Although many chatbots correctly stated that alternative therapies are not scientifically proven, some still suggested treatments such as herbal remedies, acupuncture and restrictive diets. Researchers described this as “false balance,” where scientifically unsupported options are presented alongside evidence-based medicine.

Doctors involved in the discussion warned that some alternative treatments may directly harm patients or delay life-saving medical care. Experts also reported cases where AI chatbots caused unnecessary anxiety by giving overly dramatic predictions about a patient’s survival. The researchers concluded that while AI can be useful for general information, it should not replace professional medical advice, particularly for serious conditions such as cancer or urinary health problems.

HealthDay Staff. (2026, April 21). Study finds AI chatbots can give misleading health advice. Drugs.com. https://www.drugs.com/news/study-finds-ai-chatbots-can-give-misleading-health-advice-129726.html