Each month, about 30% of adults in the US turn to artificial intelligence for health information and advice, according to a recent survey by KFF, a non-profit health policy research organization. This number has almost doubled in the last two years.
That’s no surprise, because AI allows access to fast, jargon-free advice and eliminates the need to wait for an appointment, said Dr. Kathryn Schmidt, internal medicine physician at UC Health.
But that’s not always best, she said.
“Quick and reliable talk is good, but quick and wrong can be dangerous, especially when talking about your health,” Schmidt said.
Use AI as a starting point, not a pro perspective
With busy schedules and barriers to getting prompt medical care, AI seems like an attractive, convenient way to get information, said Dylan Clark, chief analytics and AI officer at Bon Secours Mercy Health.
According to KFF, nearly 1 in 5 AI health users turned to the technology for answers because they didn’t have a health care provider or couldn’t get an appointment. Another 1 in 5 said they couldn’t afford health care.
But AI is not a safe alternative to medical care, Schmidt and Clark said.
Instead, treat AI like a “health encyclopedia,” Schmidt said. Although it’s helpful in understanding a situation, breaking down information, or preparing for an appointment, AI is not a doctor.
“Use this as a way to clarify the questions you need to ask, not to avoid seeing the doctor, delaying care, or changing the prescription yourself,” Clark said.
Here are some examples from Schmidt and Clark for feeding chatbots:
- “What do my cholesterol numbers mean?”
- “What should I ask my doctor before starting a new blood pressure medication?”
- “I’m thinking about how to adjust my lifestyle to be more healthy going forward. Do you have any suggestions for meal planning?”
Keep all findings educational rather than relying on AI for accurate diagnoses, Schmidt said.
“General health education and visit preparation is where AI really shines.”
Always check and double-check the sources used by AI
Dr. Marcus Romanello, chief medical officer of Christ Hospital Health Network, said AI can provide valuable information or answers to straightforward questions, but it lacks any oversight to validate the quality of the content generated.
For this reason, be careful with how you phrase your prompts, choosing simple, straightforward questions with direct answers, such as asking about symptoms of the diagnosed disease.
“The instruments are still prone to overconfidence in hallucinations or false responses,” Romanello said. “This makes it harder to identify dangerous advice.”
Be wary of any answer an AI gives with confidence, Clark said. AI often produces information even when it appears definitive.
Ask the chatbot to cite its sources and then verify that those sources are trustworthy, such as established health systems, academic centers, or government health agencies.
AI can also create citations that don’t exist, Schmidt said. Therefore, be sure to cross-reference any important information with reliable sites like MedlinePlus or Mayo Clinic.
Keep your personal information private
It’s important to check the information you get from AI, Schmidt and Clark said, but it’s also important to make sure you’re not giving out too many personal details.
Medical records are protected by the Health Insurance Portability and Accountability Act, or HIPAA. However, not all information you give to an AI chatbot is protected in the same way.
It should be safe to provide some details about your personal circumstances, such as your age or general lifestyle. But avoid sharing sensitive information related to your identity, such as your date of birth or address.
Never use AI in an emergency or crisis situation
There are some situations that AI can’t respond to, Clark said.
“Any urgent or big bets should get to a real person immediately,” he said.
If you are experiencing chest pain, trouble breathing, or a mental health crisis, it is time to seek care from a trained professional. In case of an emergency, call 911 or go to the nearest emergency room. During a mental health crisis, call the suicide and crisis lifeline 988.
Check with a care provider before making any medical decisions
Decisions about your medical care should be between you and a professional provider, not a chatbot. Instead of asking an AI if you should change the dosage on your prescription, talk to a doctor. Don’t act on any information given to you by the chatbot and always keep your care team informed.
“The most important thing is that a physician’s judgment should always be informed by and inform any actual decision, because AI can be wrong,” Clark said. “Its output is a prompt for conversation with a physician, never a replacement for one.”
Romanello said there are many cases in which the only way to get clear answers is through personal investigation.
If cost or access to care is a barrier, Schmidt recommends seeking out community health centers, finding telehealth services or consulting nurse advice lines.
“You matter more than convenience,” Clark said.
Reporting by Esther Lauenstein, USA TODAY Network/Cincinnati Enquirer. USA TODAY Network via Reuters Connect.