Microsoft Launches Health-Focused AI Chatbot

Microsoft is the latest technology company to roll out a health-focused artificial intelligence (AI) chatbot.

Copilot Health, launched late last week, enters a rapidly growing field that includes Amazon’s Health AI, Open AI’s ChatGPT Health, and Anthropic’s Claude for Healthcare.

As more people seek advice from chatbots, it isn’t surprising that companies are now tailoring these tools to answer health-related questions.

In an online post announcing Copilot Health, Microsoft highlighted its practical benefits for patients. “At some point, we’ve all stared at a test result we didn’t understand,” the company stated. “Worn a device that tracked everything but revealed little. Sat in a clinic waiting room with a list of questions we forgot the moment we sat down for a consultation. Felt that quiet, unsettling feeling that something is off — but had nowhere to take it.”

Targeting these issues, Copilot Health will function as a separate and secure space within Microsoft’s existing Copilot chatbot. It will draw from an individual’s shared health records, history, and wearable data to “make sense” of this information and deliver personalized insights, according to the company.

The tool “doesn’t replace your doctor,” Microsoft’s post stated. “It makes every minute you have with them count more. You arrive prepared, with the right questions, the right context, and the confidence that comes from better understanding your own body.”

Physicians have noted both potential benefits and concerns about health-focused chatbots like Copilot Health.

“I do see there’s a great need for this tool that would, with some guardrails, really allow patients to do some research, but also make it such that it doesn’t overwhelm them, doesn’t increase anxiety, and empowers them,” Misbah Keen, MD, MBI, MPH, professor of family medicine at the University of Washington School of Medicine, told MedPage Today.

“It’s all about patient empowerment,” he said. Keen is a co-founder of Keensight Health, which is developing an AI platform to support patient-centered, efficient clinical visits.

“I like to think of AI as a really fast assistant,” noted Jen Brull, MD, board chair of the American Academy of Family Physicians. “It can generate ideas in seconds, but it is not a doctor who knows you,” she told MedPage Today in an email.

“AI can do things like help spark healthy meal ideas when you’re stuck in a dinner rut; it can help patients feel more informed ahead of a doctor’s appointment and even help you research your symptoms. For some people, that convenience is genuinely helpful,” Brull pointed out. “And for physicians, there are also ongoing efforts to integrate trusted clinical research and published clinical guidance into AI tools.”

Clinicians have acknowledged that some complexities innate to health data might be simplified through chatbots, which can provide more personalized information than a typical online search, improve patient education, help patients understand data from various apps and trackers they use, and ease healthcare visits.

But some have expressed concerns. “We don’t want to create a system wherein the patient comes up with these reams of data, and [the physician doesn’t] have time to go over it. That’s a setup for disaster,” Keen noted. Patients are “not going to feel satisfied,” he added.

AI “should not replace personalized medical advice from a usual source of care,” Brull emphasized. “A family doctor or trusted primary care clinician should always be involved in health decisions. We know your history, your risks, and the full picture of your health, and that is something no chatbot can replicate.”

There are also risks of patients misinterpreting information or AI being incorrect. One Washington Post reporter recently uploaded a decade’s worth of data from his Apple watch and received an “F” in cardiac health from a chatbot. His doctor said that assessment couldn’t be further from the truth.

Another concern is that people may misuse AI tools, relying on them during a medical emergency instead of seeking professional care, for example. There are potential privacy risks as well.

Microsoft underscored that Copilot Health is “not intended to diagnose, treat, or prevent diseases or other conditions and is not a substitute for professional medical advice.”

Conversations and data are isolated from the existing Copilot chatbot and “kept under additional access, privacy, and safety controls,” the company added. Safeguards include encryption at rest as well as in transit, strict access controls, and the ability for individuals to manage and delete their information when they choose. User information will not be used for model training.

The tool also connects to healthcare directories so users can search for clinicians by specialty, location, language, and insurance coverage.

Physicians should help determine how to best use these new tools, rather than saying “we shouldn’t be using them,” Keen said.

Microsoft noted the importance of physician involvement developing Copilot Health: besides an internal clinical team, an external panel of more than 230 physicians was involved.

Last week, the company shared a link for people to sign up to be among the first Copilot Health users. It’s not clear how many have registered so far.