Before asking AI about your symptoms, read this guide
The Health Chatbot User Guide introduces 5 steps under the name SMART. It covers how to ask questions, protect personal information, check answers and seek professional help when necessary.
The guide was developed by an international group of patients, doctors, researchers, AI experts and ethicists.
The need for such guidance is growing. A March 2026 survey by the US health research organization KFF found that 32% of American adults had used AI for health information or advice in the previous year. A separate UK survey conducted for the guide found that 43.7% of adults had asked an AI chatbot a health question during the same period.
However, research shows that AI answers can be unreliable, particularly when people use them to make medical decisions.
A 2026 study published in Nature Medicine involved 1,298 participants who used AI chatbots to assess 10 medical scenarios. When researchers tested the chatbots without human users, they identified the relevant conditions correctly in 94.9% of cases.
But when ordinary people used the same chatbots, they identified relevant conditions in fewer than 34.5% of cases. The results were no better than those of participants who used their usual information sources.
Another study, published in npj Digital Medicine in February 2026, examined 888 responses from four AI chatbots to 222 patient questions. Between 5% and 13% of their answers contained unsafe advice that could potentially lead to serious harm.
A survey by King's College London also highlighted the risks. It found that 15% of UK adults had used AI for health advice instead of contacting a doctor or another health service. Among people who sought health advice from chatbots, 21% said something the chatbot told them had discouraged them from seeking professional help.
Five steps for safer AI use
The SMART guide offers practical advice for people who want to use chatbots to understand their health or prepare for a medical appointment.
S: Slow down and think. Decide what you want to learn before opening a chatbot. AI can explain unfamiliar medical terms, describe general information about a condition and help prepare questions for a doctor. It should not replace a medical examination or determine whether urgent treatment is necessary.
M: Make your questions clear. Provide relevant context, such as how long symptoms have lasted. Ask for explanations in plain language and request questions to discuss with a doctor. Avoid treating a chatbot's suggested diagnosis as a medical conclusion.
A: Ask yourself what you are sharing. Avoid entering personal details such as your full name, address, identification numbers or other information that could identify you. Be careful when uploading medical records or test results. Check the chatbot's privacy policy and data settings before sharing sensitive information.
R: Remember to check the facts. Ask the chatbot to provide links to reliable medical sources, then open and examine them yourself. Check when the information was published and whether it comes from a recognized health organization. AI can produce convincing answers while providing incorrect or irrelevant sources.
T: Talk to a real person when needed. Discuss AI-generated information with a doctor, especially when it concerns symptoms, test results or treatment. Tell the clinician that you used a chatbot so they can address any misunderstandings. Never start, stop or change prescribed medication based solely on AI advice.
The guide also stresses that people should not use chatbots instead of emergency medical services. In Kazakhstan, people who need urgent medical assistance can call an ambulance at 103 or the unified emergency number 112.
Earlier, Qazinform News Agency reported that AI could analyze electrical signals from the heart during sleep to help predict a person's risk of heart failure, irregular heartbeat, stroke and death up to 10 years ahead.