AI and health information: how to use it safely

First, the most important rule

AI does not replace a qualified doctor. A chatbot cannot examine you, does not know your full history and can be confidently wrong. Use it to understand and prepare, never to diagnose yourself or decide on treatment.

In an emergency, do not ask a chatbot. If someone has chest pain, difficulty breathing, heavy bleeding, a serious injury, sudden weakness or confusion or has swallowed something harmful, call your local emergency service or go straight to the nearest hospital emergency department.

What AI can help with

Used carefully, a chatbot can make a visit to the doctor easier and less confusing.

• Understanding a word. Your report says something you do not recognise. Ask "What does this medical word mean, in simple Urdu?"
• Preparing questions. Before an appointment, ask "I am seeing a doctor about my mother's diabetes. What questions should I ask?" Write the useful ones down and take them with you.
• Describing your symptoms clearly. Ask for help putting what you feel into a short, clear list: when it started, how often it happens and what makes it better or worse. This helps you explain it to the doctor.
• Understanding general advice. After your doctor explains something, you can ask a chatbot to explain the general idea again in simpler words.
• Healthy habits. General information about sleep, exercise or a balanced diet can be a useful start, as long as you check anything that affects a health condition with your doctor.

What AI must never replace

• A diagnosis. Many illnesses share the same symptoms. Only a qualified doctor who examines you can tell them apart.
• A decision to start, stop or change a medicine. Never do this because a chatbot suggested it.
• Emergency care. Every minute matters in an emergency. Get help first.
• Mental health support in a crisis. If you or someone you know is thinking about self-harm, contact a doctor, a hospital or someone you trust immediately.
• Advice for pregnancy, babies and children. Their needs are different. Always ask a doctor, a lady health worker or another qualified health professional.

Medicines need extra care

Medicines are where AI mistakes can do the most harm.

1. Never take a dose from a chatbot. Doses depend on age, weight, other medicines and health conditions. A chatbot can get any of these wrong.
2. Do not trust a chatbot on brand names. The same medicine is sold under different names. A chatbot may confuse similar names.
3. Read the leaflet and ask a pharmacist if you are unsure how to take a medicine.
4. Ask your doctor about mixing medicines, including herbal remedies and medicines bought without a prescription.

Keep your health information private

Health information is among the most sensitive personal data you have.

• Remove names and numbers first. Write "a woman aged 55" instead of your mother's name. Never include a CNIC number or hospital record number.
• Do not upload full reports with your name on them. If you need help with a word, type only that word or line.
• Check the app's privacy settings. Many chatbots let you delete chat history or stop your chats being used to improve the service.
• Be careful with health apps. Download apps only from the official app store. Be wary of any app that asks for more information than it needs.

Warning signs of bad health information

• A promise of a guaranteed cure
• Advice to stop a medicine your doctor prescribed
• Pressure to buy a product, supplement or treatment
• Confident answers with no mention of seeing a doctor
• Claims shared on WhatsApp or social media with no clear source

If you see these, stop and speak to a qualified doctor.

What to check before you act

• Has a qualified doctor or pharmacist confirmed this?
• Did I remove names and personal details before asking?
• Am I using AI to understand, not to decide?

To understand why chatbots can sound so sure when they are wrong, read How chatbots work.