������Ƶ

More people are using AI for medical advice. Here what to know

AI can be useful to physicians and patients alike — but its answers are not always reliable. Photo credit: SLuisa P Oswalt/Shutterstock.

by Nate Miller, Bridge Michigan

Type health symptoms into your phone or laptop, and one of numerous artificial intelligence models will spit out potential diagnoses and treatments in seconds.

But Dr. ChatGPT isn’t ready to replace your family physician just yet.

Nearly 1 out of every 3 adults reported asking artificial intelligence for health advice, . And while some doctors who spoke to Bridge Michigan acknowledge their office visits are more efficient when patients learn about their conditions online, there is also a chance Grok or Gemini is steering them wrong.

Even a query to Google AI mode admits the limitations of AI medical consultations.

“AI medical advice is not consistently accurate and should not be relied upon for diagnosis or treatment,” Google AI mode says. “While generative AI models score highly on standardized medical exams, recent real-world studies show that AI chatbots provide incorrect or poor medical answers up to 24% to 50% of the time.”

Here what you need to know:

How has AI changed doctor visits?

Doctor consultations are often quite short, and AI has made those minutes more efficient for patients and physicians.

Medical literature is notoriously hard to understand for people without medical training. AI can quickly reduce dense medical research into simple answers.

Dr. Latisha Malcom, who has been a primary care doctor for 29 years, said AI has greatly improved the communication she has with patients, because it can explain their symptoms in terms they can understand.

“If you’re seeing a bunch of symptoms and you don’t know how to put it together, and you put it all in, the AI system has given you a possibility of what it could be,” said Malcom, who is the chief medical officer of . “I think that’s great, and we can put that together and see what we can come up with, and at least start a workup of those things.”

Is there a downside to asking AI about medical concerns?

AI offers quick and simple answers, but those answers can be wrong.

Nearly half of medical answers generated by AI chatbots are problematic, according to a . In a separate study, of doctors reported having to intervene after a patient had followed through with unsafe online health information, including information from AI.

“In most cases AI has identified something that maybe fits generally within (the patient) symptoms, but importantly is excluded for other reasons that maybe weren’t entered into the system or just not appreciated by AI,” said Dr. Bradley Uren, vice chair of the Board of Directors for the . “AI is not performing the examination. It doesn’t necessarily have all of the data.”

How can I best use AI to prepare for a doctor visit?

AI is most useful as a tool to organize symptoms. It allows patients without much health literacy to present them to their doctors in a coherent way, Uren said. “Often the appointment times are just a few minutes, so time is precious, and that can help.”

Doctors report that their patients who use AI are often led to scary diagnoses.

“If there is a constellation of symptoms that the machine is evaluating, it may make sense in some cases for it to jump to the most serious conclusion rather than the most benign conclusion,” said Uren.

Being specific in the information you provide AI may help with that.

“If you narrow it down and give more details about your particular age, your prior diagnoses and your other (lab results) … you can get a much more specific, nuanced answer,” said Malcom.

AI is best used as a tool to start doctor-patient dialogue, not as a final medical authority.

Can AI replace a doctor visit?

Because AI can respond to a specific set of symptoms with potential diagnoses, and do so with an authoritative voice, it can be particularly dangerous to rely on.

“There’s no medical authority backing it up,” said Dr. Rachel Klamo, former president of the . “It can sound like it has medical authority, and I think it can convince people that it does. But at the end of the day, the people who have actually trained and have been certified to diagnose are going to be the ultimate authorities.

“A lot of the AI tools, they’re generated around the idea of, ‘Get a quick, simple, perfect answer,’” Klamo said. “That’s rarely the way things are in medicine.”

If a patient is convinced that AI has diagnosed him or her correctly and his or her doctor has not, precious clinician time may be wasted convincing patients of the correct treatment path.

“Unfortunately, that is part of the job,” said Klamo. “This tool has a lot of ways that it can be used for good … but it has a lot of potential downsides.”

Is my health information private on AI?

Chatbots that people use on a daily basis, like ChatGPT, Gemini and Claude, are not bound by the same laws protecting private health information as are health professionals. Patients who share health data with an AI chatbot are essentially making that information available to whatever company runs the chatbot.

“A lot of the tools that are out there are not necessarily protecting (patients’) interests in terms of their health information,” said Jodyn Platt, associate professor of learning health sciences at the .

While it can be tempting to ask AI to explain a symptom, test result or medical document, said Platt, when you do, “you’ve just basically shared your private health information with OpenAI.”

Is my doctor using AI?

More than 80% of doctors use AI professionally, according to a . Doctors use it for multiple tasks — they may use ambient listening technology, which generates charts based on audio recordings of their interactions with patients, consult a chatbot about treatment plans or simplify administrative work.

“We were always trying to give good clinical guidelines and things of that nature throughout the course of my practice,” said Malcom. “Now I have it at my fingertips, and that makes it much quicker for everyone.”

AI can do more, though — a model outperformed doctors on a variety of clinical tasks including diagnoses, according to a .

But today, to supplement the care they provide, not to override it.

“A lot of physicians have taken to calling this ‘augmented intelligence,’ rather than the sort of traditional artificial intelligence,” said Uren. “It’s augmenting the intelligence of the clinician to help collate data or help rapidly search vast volumes of data to find a few needles in that haystack that then can be more closely evaluated.”

AI isn’t going anywhere — so what does the future look like?

While AI could plausibly replace some human doctors, experts don’t think it should.

“I think we really need people,” said Platt. “I think it will change the way we work and the way we relate to one another, but I don’t see a reduction in workforce all the way around. My hope is that it doesn’t go that direction.”

If it does, it may become increasingly difficult to access human health professionals.

“The nightmare scenario is that you end up having health care systems that operate like Comcast or Xfinity, where it’s impossible to get to a human being that you really need, and I think health care shouldn’t go that way,” said Platt.

For now, Uren said he is “unaware of an AI system where I would tell patients that this is a source of information that you can trust.

“The trusted source for (patients’) health care information should be their health care professional,” Uren said.

This first appeared on and is republished here under a .

by Sara Swanson

Sara Swanson is an editor, writer and a founder of the Manchester Mirror.