A few years ago, I thought I was about to die. And while (spoiler alert) I didn’t, my severe health anxiety and ability to always assume the worst has persisted. But the increasing proliferation of health tracking smart devices and new ways that AI tries to make sense of our body’s data has led me to make a decision. For my own peace of mind, AI needs to stay far away from my personal health. And having just watched Samsung’s Unpacked event, I’m more convinced of this than ever. I’ll explain.

Sometime around 2016, I had severe migraines that persisted for a couple of weeks. My anxiety steeply increased during this period due to the attendant worry, and when I eventually called the UK’s NHS helpline and explained my various symptoms, they told me I needed to go to the nearest hospital and be seen within 2 hours. “Walk there with someone,” I distinctly remember them telling me, “It’ll be quicker than getting an ambulance to you.”

This call confirmed my worst fears — that death was imminent. 

As it turned out, my fears of an early demise were unfounded. The cause was actually severe muscle strain from having hung multiple heavy cameras around my neck for an entire day while photographing a wedding. But the helpline agent was simply working on the limited data I’d provided, and as a result, they’d — probably quite rightly — taken a “better safe than sorry” approach and urged me to seek immediate medical attention. 

I’ve spent most of my adult life struggling with health anxiety, and episodes such as this have taught me a lot about my ability to jump to the absolute worst conclusions despite there being no real evidence to support them. A ringing in my ears? Must be a brain tumor. A twinge in my stomach? Well, better get my affairs in order. 

I’ve learned to live with this over the years, and while I still have my ups and downs, I know better about what triggers things. For one, I learned never to Google my symptoms. Because no matter what my symptom was, cancer was always one of the possibilities a search would throw up. Medical sites — including the NHS’s own website — provided no comfort and usually only resulted in brain-shattering panic attacks. 

Sadly, I’ve found I have a similar response with many health-tracking tools. I liked my Apple Watch at first, and its ability to read my heart rate during workouts was helpful. Then I found I was checking it increasingly more often throughout the day. Then the doubt crept in: “Why is my heart rate high when I’m just sitting down? Is that normal? I’ll try again in 5 minutes.” When, inevitably, it wasn’t different (or it was worse), panic would naturally ensue. 

Whether tracking heart rate, blood oxygen levels or even sleep scores, I’d obsess over what a “normal” range should be and any time my data fell outside of that range, I’d immediately assume it meant I was about to keel over right there and then. The more data these devices provided, the more things I felt I had to worry about. I’ve learned to keep my worries at bay and have continued to use smartwatches, without them being much of a problem for my mental health (I have to actively not use any heart-related functions like ECGs), but AI-based health tools scare me. 

During its Unpacked keynote, Samsung talked about how its new Galaxy AI tools — and Google’s Gemini AI — will supposedly help us in our daily lives. Samsung Health’s algorithms will track your heart rate as it fluctuates throughout the day, notifying you of changes. It will offer personalized insights from your diet and exercise to help with cardiovascular health and you can even ask the AI agent questions related to your health.

To many it may sound like a great holistic view of your health, but not to me. To me it sounds like more data being collected and waved in front of me, forcing me to acknowledge it and creating an endless feedback loop of obsession, worry and, inevitably, panic. But it’s the AI questions that are the biggest red flag for me. AI tools by their nature have to make “best guess” answers based usually on information publicly available online. Asking AI a question is really just a quick way of running a Google search, and as I’ve found, Googling health queries does not end well for me. 

Much like the NHS phone operator who inadvertently caused me to panic about dying, an AI-based health assistant will be able to provide answers based only on the limited information it has about me. Asking a question about my heart health could bring up a variety of information, just as looking on a health website would about why I have a headache. But much like how a headache can technically be a symptom of cancer, it’s also much more likely to be a muscular twinge. Or I haven’t drank enough water. Or I need to look away from my screen for a bit. Or I shouldn’t have stayed up until 2 a.m. playing Yakuza: Infinite Wealth. Or a hundred other reasons, all of which are far more likely than the one I’ve already decided is definitely the culprit. 

But will an AI give me the context I need to not worry and obsess? Or will it just provide me with all the potentials as a way of trying to give a full understanding but instead feeding that “what if” worry? And, like how Google’s AI Overviews told people to eat glue on pizza, will an AI health tool simply scour the internet and provide me with a hash of an answer, with inaccurate inferences that could tip my anxiety into full panic attack territory? 

Or perhaps, much like the kind doctor at the hospital that day who smiled gently at the sobbing man sitting opposite who’d already drafted a goodbye note to his family on his phone in the waiting room, an AI tool might be able to see that data and simply say, “You’re fine, Andy, stop worrying and go to sleep.” 

Maybe one day that’ll be the case. Maybe health tracking tools and AI insights will be able to offer me a much-needed dose of logic and reassurance to counter my anxiety, rather than being the cause of it. But until then, it’s not a risk I’m willing to take.