AI chatbots give risky cancer treatment advice

A study published in BMJ Open examined how leading AI chatbots (including ChatGPT, Gemini, Meta AI and others) respond to queries about cancer treatment. Researchers found that nearly half of the AI-generated answers were problematic. About 30 % were rated “somewhat problematic,” containing partially accurate but incomplete information, while roughly 20 % were deemed “highly problematic,” providing outright incorrect guidance. In addition to stating that standard chemotherapy remains essential, the bots also suggested unproven alternatives such as acupuncture, herbal remedies, and specialized “diet against cancer” plans, and even listed clinics that focus solely on alternative therapies.

Experts warned that reliance on chatbots for health advice could amplify misinformation, especially as more patients turn to these tools as a primary source of medical information. The authors emphasized the need for public education and stricter oversight to prevent dangerous health outcomes.