You can. But AI may agree with you when it should correct you.
By Elizabeth Ndungu | Founder, Ndungu Consulting | Tech Coach for Adults 50+
You receive your blood test results. The letter uses words you do not recognize, and your appointment is three weeks away. So you ask an AI tool what the results mean.
Many people do this now because AI is quick, available at any hour, and easier to read than most medical paperwork. A 2026 KFF survey found that nearly one in three U.S. adults had used an AI chatbot for health information.
AI can explain general terms and help you prepare for an appointment. It should not make medical decisions for you.
AI Health Answers Can Sound Better Than They Are
In April 2026, researchers published an audit in BMJ Open. They tested five popular chatbots with questions about cancer, vaccines, nutrition, stem cells, and athletic performance.
Nearly half of the answers were rated problematic.
Some lacked context. Others contained misleading information or serious errors. They still sounded polished and professional, which made the problems harder to spot.
Earlier research has found serious problems with AI-generated references, including incomplete, inaccurate, and fabricated citations. Newer models perform better, but any source an AI provides still needs to be checked.
An answer is not reliable simply because it sounds confident and cites sources.
AI May Agree With You Too Easily
Researchers call this sycophancy. It means the chatbot may support your assumption instead of challenging it.
This question already contains a conclusion:
Why does taking a high dose of vitamin C protect me from cancer?
A more useful version gives the tool room to correct you:
Is there reliable medical evidence that high doses of vitamin C prevent or treat cancer?
A 2026 Oxford study, published in Nature, found that chatbots trained to sound warmer were around 40 percent more likely to reinforce incorrect beliefs. Warmer models also made more errors overall, up to 30 percent more on safety-critical topics like medical advice.
When users expressed sadness or emotional distress, the accuracy gap became wider. That matters because many people ask AI about health when they are already frightened or anxious.
A friendly answer can feel reassuring. It can still be wrong.
Better Ways to Ask AI About Health
Do not ask AI to diagnose you.
Instead of:
Is this chest pain just anxiety?
Ask:
What are some possible causes of chest pain, and which signs need urgent care?
You can also ask:
What information would a doctor need before giving personal advice?
Or:
What could make this answer wrong or incomplete?
Be more cautious when the answer agrees with you immediately. Easy agreement can be a warning sign, not reassurance.
Questions for AI vs Questions for Your Doctor
Ask AI: What does atrial fibrillation mean?
Ask your doctor: Do I have it, and what does it mean for my health?
Ask AI: What is metformin generally used for?
Ask your doctor or pharmacist: Is it safe with my other medications?
Ask AI: What are the general signs of low vitamin B12?
Ask your doctor: My B12 result is 189. What does that mean for me?
AI can explain general information. Your doctor can apply it to your test results, medical history, symptoms, and medications.
Do Not Use AI for Emergencies or Medication Changes
Never use AI to decide:
• If a symptom is serious
• If you should delay medical care
• If you should start or stop a medication
• If you should change a dose
- If two medications are safe together
A 2026 Mount Sinai study found that ChatGPT Health under-triaged more than half of simulated medical emergencies in structured testing. In some cases, it directed serious situations toward routine follow-up instead of urgent care.
If you have severe chest pain, trouble breathing, signs of a stroke, sudden confusion, heavy bleeding, or rapidly worsening symptoms, call 911 in the United States or your local emergency number.
Do not wait for a chatbot to reassure you.
Protect Your Medical Information
In the same KFF survey, 41 percent of people who used AI for health said they had uploaded test results or doctors’ notes. Privacy concerns remained high. Even among those who had shared medical information, 65 percent said they were worried about privacy.
Before uploading anything, remove details such as:
• Your full name
• Address
• Date of birth
• Medical record number
• Insurance information
• Prescription numbers
In most cases, you can type the medical term or sentence you want explained instead of uploading the full document.
Linda’s Story
One client, whom I will call Linda, had been asking AI about her symptoms late at night. The chatbot gave her long lists of possible conditions, including rare and serious ones.
The answers did not calm her. They gave her more to worry about.
She made one change. She stopped asking health questions at night. If an answer increased her anxiety, she closed the app and wrote the question down for her doctor.
She still uses AI to explain medical terms and prepare questions before appointments. She treats it as background reading (research purposes) , not a diagnosis.
Linda’s name and identifying details have been changed for privacy.
Use AI for the Conversation, Not Instead of It
Use AI to help you understand the conversation.
Do not use it to replace the conversation.
AI can explain medical language and help you prepare for an appointment. When the question becomes personal, such as “Do I have this?”, “Should I take this?” or “Is this an emergency?”, speak to a qualified medical professional.
A confident answer is not always correct. A warm answer is not always safe.
I write about technology safety and digital literacy for adults over 50. Follow me here on Medium for future articles.
How to Use AI: A Guide for 50+ is available here:
https://elizabethw2.gumroad.com/l/howtouseai50plus
For one-on-one technology help, visit ndunguconsulting.com.
About the Author
Elizabeth Ndungu is the founder of Ndungu Consulting, a technology coaching and digital literacy practice that helps adults over 50 build confidence with computers, phones, AI tools, email, Microsoft Office, and online safety.
Sources:
• KFF: 1 in 3 Adults Are Turning to AI Chatbots for Health Information, March 2026: https://www.kff.org/health-information-trust/poll-1-in-3-adults-are-turning-to-ai-chatbots-for-health-information-equaling-the-share-who-use-social-media-for-health/
• BMJ Open: Generative AI-Driven Chatbots and Medical Misinformation, April 2026 (coverage via CIDRAP): https://www.cidrap.umn.edu/misc-emerging-topics/ai-chatbots-provide-poor-answers-medical-questions-half-time-study-finds
• University of Oxford: Friendly AI chatbots make more mistakes and tell people what they want to hear, April 2026 (peer-reviewed in Nature 652:1159–1165, DOI 10.1038/s41586–026–10410–0): https://www.ox.ac.uk/news/2026-04-29-friendly-ai-chatbots-make-more-mistakes-and-tell-people-what-they-want-to-hear
• KFF Monitor: Companies Expand AI Health Offerings Even as Accuracy Questions Remain (Mount Sinai under-triage finding), April 2026: https://www.kff.org/health-information-trust/companies-expand-ai-health-offerings-even-as-accuracy-questions-remain/
Tags: Artificial Intelligence · Health · Adults Over 50 · Digital Literacy · Technology
Leave a Reply