AARP Hearing Center
More People Are Turning to AI for Health Advice — Should They?
AI chatbots are good at things like interpreting medical jargon and helping patients prepare for appointments, but experts caution against relying on them for diagnoses and treatments
AARP (Getty Images)
Before Ruth Furman was diagnosed with endometrial cancer last year, she avoided asking AI for health advice to protect her anxiety-prone personality from spiraling. “I would ask, ‘I have a leaky showerhead, what should I do first?’ ” the 59-year-old Las Vegas publicist told AARP. “But if my body were leaking, I would not ask AI.”
Even after her diagnosis and during treatment, Furman largely stuck to her policy. But once she reached survivorship, where she was monitored every few months, she occasionally turned to chatbots when facing fears of recurrence. It mostly confirmed her initial hesitation.
More Ways to Benefit
- Related Benefit 1
- Related Benefit 2
- Related Benefit 3
- Related Benefit 4
- Related Benefit 5
“I have had what I thought were symptoms [of cancer], and I got nervous and consulted Chat,” Furman says. But doing so, she said, only made her more nervous. Now she just calls the doctor’s office instead.
Furman is hardly alone. As AI tools become more widely available, more people are turning to them for quick health answers. A recent poll from the health policy nonprofit KFF found that about a third (32 percent) of adults turn to AI for health information and advice — and the majority are doing so for quick and immediate answers.
And while younger adults are among those most likely to use AI for health advice, older adults are doing it too. Roughly 2 in 5 adults over age 50 who use AI (41 percent) have already asked an AI-powered tool a health-related question, according to a new AARP survey. In the next year, even more 50-plus AI users (62 percent) say they are likely to seek answers from AI to their health questions.
Some, too, may already be doing so without fully realizing it: Search engines like Google automatically generate AI overviews to queries.
“The AI chatbots, online health assistance — all these things are very highly, I find, in the forefront of many older people’s minds,” says Dr. Jeffrey Kaye, a professor of neurology at Oregon Health & Science University, where he directs the Oregon Center for Aging and Technology.
Turning to AI for health advice can be helpful in some cases, experts say — but it also comes with risks. Here’s what to know before you type in a health question.
AI’s limitations: It’s not ‘a stand-alone clinician’
More than 40 million people turn to ChatGPT every day with health questions, according to its maker, OpenAI. But studies suggest that in their current form, AI chatbots can fall short in meaningful ways — especially when it comes to health.
For instance, they struggle when reasoning is needed. A study led by Mass General Brigham researchers asked 21 AI tools, including ChatGPT, Claude and Gemini, to essentially “play doctor.” The results, published in April in JAMA Network Open, showed that the tools failed 80 percent of the time at earlier stages of clinical decision-making, such as deciding which diagnoses to consider before enough information is available to confirm one.
This is concerning because it’s the stage when patients are most likely to turn to AI — when they first notice a symptom or sign of illness, according to study coauthor Dr. Marc Succi. In fact, more than half (55 percent) of people who use ChatGPT for health questions use it to check or explore symptoms.
However, once the AI models had all the relevant data, they were better at landing on a diagnosis. In other words, Succi says, “it’s not hard to do well when you have an open-book test, which is how these tools work.” The results, he adds, suggest that AI has the potential to help physicians but not replace them.
“For patients, especially older adults with multiple conditions or medications, the takeaway is not to not use AI; it’s to treat AI not like a stand-alone clinician and always get some human eyes on the output,” says Succi, a radiologist at Massachusetts General Hospital and executive director of the MESH Incubator.
Another paper, published in Nature Medicine, found that people who turned to an AI chatbot with health questions identified the likely condition only about one-third of the time. What’s more, the researchers found that the bots often provided a mix of good and bad information, making it hard to know the best next step.
AI chatbots can also underrecognize health emergencies. A research team at the Icahn School of Medicine at Mount Sinai created 60 clinical scenarios that ranged in levels of urgency, from cases that could be treated at home to those that required emergency care. The study, published in the journal Nature Medicine, found that ChatGPT Health failed to recognize the seriousness of more than half (51.6 percent) of the medical emergency scenarios tested.
“ChatGPT Health performed well in textbook emergencies such as stroke or severe allergic reactions,” lead author Dr. Ashwin Ramaswamy explained in a news release. “But it struggled in more nuanced situations where the danger is not immediately obvious, and those are often the cases where clinical judgment matters most. In one asthma scenario, for example, the system identified early warning signs of respiratory failure in its explanation but still advised waiting rather than seeking emergency treatment.”
How to use AI smartly for health advice
That’s not to say AI can’t help you with your health; experts interviewed for this article say it can be beneficial in several ways. “It just all comes down to how it’s used,” Succi says, “and to manage your expectations of the quality of the results.”
Here are a five best practices for using AI in health care:
1. Use it to become a more informed patient. Instead of using AI to become your own doctor, use it to become a better patient, Succi says. For example, ask it to help you prepare questions for your next medical appointment — something Furman says she does — or outline concerns you want to share with your health care provider. AI can also be helpful in translating medical jargon, like from test results, into plain English.
And don’t shy away from bringing what you’ve discovered to your doctor. Kaye says when patients come into his clinic with printouts of health information they found online, he sees it as a positive sign. “Any practitioner should be glad that the person is trying to educate themselves,” he says.
2. Give enough context — but protect your privacy. Older adults, especially, are more likely to have multiple health conditions and take multiple medications. So if you’re going to type a health question into AI, it’s important to include all the relevant information.
“The issue is feeding the right information and complete information to the chatbot because it may say something that might assume you have no other medical problems or you’re not taking a certain medication that might interact with something it recommends,” Succi says.
That said, “this data goes into the database, the cloud, for these large language models to use,” Kaye says, “so you are putting your personal health information … out to the world.” For this reason, many experts warn against sharing personal information, like medical records, with AI platforms.
Turning off settings that allow a chatbot to train on your data is another step you can take to help protect your privacy, though it’s important to note that doing so doesn’t mean the companies won’t retain your data for other purposes.
“You should assume that anything you put in a chatbot from a corporation is not private anymore,” Succi says.
Most older adults are heeding that advice, AARP’s most recent survey found. In it, 69 percent of respondents said they do not feel comfortable sharing their personal health information with an AI chatbot or agent. Survey respondents also expressed concerns about their information being sold or shared without their consent, as well as privacy or security issues related to their health data.
3. Always check with a doctor — especially if you are 50-plus. In older adults, especially, symptoms for a range of illnesses “can often be less textbook,” Succi says, “and missing a diagnosis could be much higher stakes.”
What’s more, a chatbot might not fully understand the complexities of cognitive changes, caregiver concerns or how several chronic health conditions may interact, Succi says. “And so while it can offer plausible explanations, you really want a doctor to examine you, to check your vital signs, look at your medications, et cetera,” he adds.
Dr. Sid Khurana, a Las Vegas–based psychiatrist and cofounder of Psychiatry Connection and Nevada Mental Health, agrees. “It is truly important to connect to a human, a professional, before making any decisions about whether you’re OK or you’re not OK,” he says.
4. Verify the source of any answer. A good rule of thumb: If you ask AI a health question, always ask it to provide a source, then check the source of the answer, Khurana says.
“If the source is a published journal that is peer reviewed, the information is likely to be good versus if it is [from] a forum on Google — that is no different than you would have just searched yourself on Google, and we generally tell people to be careful because now you’re taking the advice of strangers,” Khurana explains. “It’s very hard to say, ‘Believe AI answers’ or ‘Don’t believe AI answers.’ I would say, look into the source of the answer.”
It’s also important to consider confirmation bias, or the tendency to favor information that supports what you already believe. A study published in the journal Nature found that so-called warmer, or more empathetic, chatbots are around 40 percent more likely to agree with a user’s incorrect beliefs. In other words, chatbots may reinforce misinformation, instead of correcting it.
That tendency can be especially concerning when it comes to health misinformation. In another study, published in the journal BMJ, researchers prompted five popular chatbots with 10 health questions from five categories, including cancer, vaccines, stem cells, nutrition and athletic performance. The researchers used prompts designed to push models toward misinformation or contraindicated advice. They found that the AI-generated responses to the questions were problematic nearly half the time.
“This behavioral limitation means that chatbots can reproduce authoritative-sounding but potentially flawed responses,” the researchers wrote.
Khurana says you can start noticing whether a chatbot is catering to your beliefs by cross-checking its responses against another AI platform. Just like news outlets spinning their reports to appeal to the political leanings of their audiences, ask yourself, Is it filtering and giving me the content that makes me want to come back and get more of it because I like what I’m getting? Khurana says.
5. Don’t let AI replace human support. When Furman did consult an AI chatbot during her cancer journey, she learned it was best to ask it what she shouldn’t do, rather than what she should. After surgery, for example, she traveled for a bridal shower and wanted reassurance that she wasn’t overdoing it or jeopardizing her recovery, even though her doctors had cleared the trip.
While Furman found that tactic helpful, the support that has mattered most over the past year hasn’t come from AI or physicians. Instead, she says, it has come from connecting with other cancer survivors in her area — something she sought out more recently because of her fear of recurrence. “This is just something that’s helping me — expanding my network of other survivors,” she says.
AARP Membership
Join AARP for only $15 per year with automatic renewal. Get instant access to members-only products and hundreds of benefits, a free second membership, and a subscription to AARP The Magazine.
More From AARP
AI's Role in Supporting Mental Health
Chatbots can help support mental health, experts say
AI Tools: Navigating Systems with Ease
AI tools streamline customer service tasks
AI Unlocks New Uses for Old Medications
The technology is repurposing existing medications to create new treatments for a variety of diseases