Heather Maxwell wasn’t feeling well. It was 2023 when the then-50-year-old’s health started to decline – something wasn’t right; she didn’t feel like herself. Brain fog, chronic fatigue, poor concentration and tightness in the chest became the norm. She’d been to doctors and underwent preliminary tests, but she was met with uncertainty. One day, not sure where else to turn and waiting for appointments with specialists, she did what most of us would be inclined to do – she turned to Dr. Google, which pointed to a slew of scary diagnoses. Filled with more questions than answers, she took her research one step further and asked ChatGPT to narrow it down. Maxwell typed in her symptoms, requested a diagnosis and waited for a reply. The super-advanced chatbot had a bunch of ideas. Now empowered with more information, she started seeing specialists who ordered some of the same tests ChatGPT had suggested, including a sleep study, which confirmed sleep apnea, explaining some of her chronic fatigue. 

Maxwell was officially diagnosed with multiple sclerosis (MS) in December 2024 – it was one of the neurological disorders the chatbot spat out. “After, I remember going back and searching ChatGPT for things like what the diagnosis meant long term, what my treatment options looked like, potential side effects, work and disability concerns, symptoms that might show up later, and suggestions on how to break the news to my family and friends,” she says, adding ChatGPT allowed her to break her new reality down into more manageable pieces. “It really helped me get a better understanding of possible outcomes.”  

This story is certainly not unique. Google has long been used to search symptoms; about half of Canadians were doing it regularly by 2013. When the pandemic rolled around, nearly 70 percent of us were using it to self-diagnose, and a quarter were using online sources to track their health. A study released in 2024 by Maple, one of the county’s biggest privately owned virtual health-care platforms, found 77 percent of Canadians have turned to search engines to look up serious symptoms instead of seeing their doctor or nurse practitioner. 

Search engines are just the tip of the iceberg. If you’ve worn a smartwatch or smart ring to track your sleep, count your steps or monitor your heart rate, or if you’ve tracked your mood, blood sugar or period on an app, you’ve used AI to gain insight into your health. So many of us use these devices and tools, but most Canadians are still unsure about the benefits and limitations of AI in health care. One reason, of course, is because all of this is still so new – it’s still the wild west out there when it comes to what using AI for health actually means. The 2024 Canadian Digital Health Survey, released by a nonprofit funded by the federal government, found only 28 percent of respondents said they’re comfortable with AI, and only a quarter felt they were knowledgeable about how it’s used when it comes to our health. 

Today, AI is behind robotic surgeries, drug development, analyzing diagnostic tests, telemedicine in rural areas, enhancing personalized medicine and reducing the administrative burden on doctors and nurses. Still, fear prevails. “In some areas, people can see and feel the technologies’ control, and they decide when to engage. AI in health is different. It’s life-and-death matters; the stakes are higher,” says Dr. Samira A. Rahimi, the Canada Research Chair in AI and Advanced Digital Primary Health Care, and an assistant professor at Montreal’s McGill University and Mila-Quebec AI Institute. “Older adults, in particular, often express concern about loss of human connection and privacy. Health care has always relied on trust, the human touch of a clinician, the confidentiality of a medical record. When AI enters the equation, it can feel like that confidential space is being shared with an invisible, unaccountable ‘third party.’” 

What we need now, says Dr. Rahimi, is to educate ourselves on the positives and negatives AI has. “When people see AI not as a replacement for clinicians but as a partner – one that analyzes scans faster, predicts risk more accurately and frees doctors’ time for more patient interaction – attitudes could shift,” she says. 

The Good

One of the biggest boons is how AI is helping health-care practitioners with their record-keeping, giving them more time and bandwidth for patients. For example, Dr. Ross Mitchell, Alberta Health Services Chair in AI in Health and a professor at the University of Alberta, and his team developed an AI scribe dubbed “Jenkins,” which automates the process of creating medical notes from practitioners’ appointments with patients. “Physicians hit record, then they can turn and focus on the patient instead of typing and glancing over every 30 seconds. They feel this makes them better physicians because they get visual cues and feedback from the patient that guides their conversations,” he says. “The patient is also much happier with the encounter because they feel like they’ve been listened to.”

Some still worry about privacy; others worry about the potential for mistakes. When it comes to the former, Dr. Mitchell points to confidentiality and security. His model, for example, doesn’t care about patients’ names or locations – it cares about medical history, so it can predict what might be coming down the line. What’s more, he says research is done in what’s referred to as “sensitive data research environments” and records are kept safely behind Alberta Health Services’ firewall. “It’s the same level of security that the banking and defense industries use.” And if you’re worried about where the data collected by AI goes, Dr. Muhammad Mamdani, the director of the University of Toronto Temerty Faculty of Medicine Centre for Artificial Intelligence Research and Education in Medicine, says it’s fair to inquire (though many physicians won’t know), but we should be practical, too. “To me, it’s far more concerning that we’re giving apps on our smartphones permission to track us without a second thought,” he says. “Everyone has to decide for themselves how comfortable they are in terms of privacy. We can complain about our data going to the government, but we’re all giving it freely to the private sector every day.” 

As for mistakes in the AI notes, Dr. Mitchell says we can’t assume humans are perfect, nor can we assume Jenkins is the perfect listener. Still, the error rate in his model is incredibly low, and physicians must review their notes to make necessary edits prior to finalizing files. “So, if AI makes an error five percent of the time, and physicians are making errors 10 percent of the time, we should only see errors one percent of the time when AI and humans work together.”

Another example that’s benefitting both practitioners and patients is how AI is playing a role in the mental health space – an area we know is in demand and underserved. Dr. Mamdani says a recent trial showed chatbot therapists perform about as well as human therapists. “We don’t have enough psychiatrists, psychologists and other mental health experts; people are sometimes waiting years to see someone. AI enables us to help the masses – it democratizes and liberates knowledge. If AI can help, that’s much better than people suffering without any hope of support,” he says. 

The Not-So-Good

Total accuracy isn’t guaranteed, and according to Dr. Rahimi, that’s largely because of “data that’s incomplete, biased or unrepresentative of certain populations, such as older adults, women, racialized groups or people with complex chronic conditions, like kidney disease.” Besides that, AI platforms (think ChatGPT, Perplexity and Grok) can only really offer general advice because it doesn’t “know” a patient’s medical history or lab results, so you can’t really have blind trust in it. Even OpenAI CEO Sam Altman has warned about AI “hallucinations,” where models generate believable info that ends up being totally incorrect. (He’s even said AI is prone to “making stuff up.”) “Without context and diverse data, even the most advanced AI system can oversimplify or misinterpret complex conditions. That’s why these tools should always be viewed as assistants,” she says. “They can guide learning, improve communication and increase engagement, but the ultimate decisions – especially in health care – must remain in human hands.” 

Zoomer Rx

More Canadians are using health care chatbots – Ada Health, Symptoma, Buoy Health and Bonjour-santé are a few of the popular ones – for basic knowledge about a personal or family member’s diagnosis, as well as to help interpret medical tests and notes from physicians. The experts say these are smart ways to try AI and see how it can help you.   

Maxwell is still using ChatGPT nearly a year after her diagnosis, specifically to learn about the progression of her disease. “I often use the prompt, ‘I’m feeling like X today – could this be MS-related?’” she says. Still, she always remembers to take what she reads with a grain of salt, and she’s learned how to make AI work for her. “ChatGPT helps me make the most of my time with my care team – it helps me get educated, prepare questions and allows me to continue my research following appointments. It’s valuable for processing information at my own pace, which ultimately helps me feel more confident in making decisions.”