Jun 29, 2025

Comments Drawer I asked AI for insight into my health data. It gave me a regurgitated book report. Link Threads | Image: The Verge, Getty Images I asked AI for insight into my health data. It gave me a regurgitated book report. Link Threads Victoria Song is a senior reporter focusing on wearables, health tech, and more with 13 years of experience. Before coming to The Verge, she worked for Gizmodo and PC Magazine. After nearly a decade of wearables testing, I’ve amassed a truly terrifying amount of health and fitness data. And while I enjoy poring over my daily data, there’s one part I’ve come to loathe: AI summaries. Over the last two years, a deluge of AI-generated summaries has been sprinkled into every fitness, wellness, and wearable app. Strava introduced a feature called Athlete Intelligence , pitched as AI taking your raw workout data and relaying it to you in “plain English.” Whoop has Whoop Coach , an AI chatbot that gives you a “Daily Outlook” report summarizing the weather, your recent activity and recovery metrics, and workout suggestions. Oura added Oura Advisor , another chatbot that summarizes data and pulls out long-term trends. Even my bed greets me with summaries every morning of how its AI helped keep me asleep every night. Each platform’s AI has its nuances, but the typical morning summary goes a bit like this: Good morning! You slept 7 hours last night with a resting heart rate of 60 bpm. That’s in line with your weekly average, but your slightly elevated heart rate suggests you may not be fully recovered. If you feel tired, try going to bed earlier tonight. Health is all about balance! Related That might seem helpful, but those summaries are usually placed next to a chart with the same data. It’s worse for workouts. Here’s one that Strava’s Athlete Intelligence generated for a recent run: Intense run with high heart rate zones, pushing into anaerobic territory and logging a relative effort well above your typical range. Thanks? I can ask Athlete Intelligence to “say more,” but it regurgitates my effort, heart rate zone, and pace metrics I can see in graphs in the workout summary. If you didn’t know anything about my athletic history or the circumstances surrounding this run, this summary might read as insightful. Here’s what the summary left out: It was dangerous to triple my mileage in only my second run of the year, given the high humidity, 85-degree-plus weather, and my spotty workout history over the past two months compared to the six months before it. Strava has access to weather data and every workout I’ve done in the past five years. I had to cut this run short because I fell and shredded my hand and knees. This is information Strava has access to, as I uploaded a gnarly picture in addition to text notes. After adding said notes, the updated summary only reflected that I cut the run short. My injury changed nothing about its insights, even though it’s the most important thing that happened during this run. I don’t know guys, without Athlete Intelligence, would I have known my elevation gain was a modest 88ft? Screenshot: Strava A more helpful insight might’ve been: “You ran during record-breaking heat for your region. While you maintained a consistent and steady pace, you have a bad habit of ramping up mileage too quickly after prolonged breaks, leading to several self-reported injuries in the past five years. A safer alternative would be lower mileage runs over two weeks to acclimate to rising temperatures. Since you’re injured, stick to low-intensity walks until your wounds have healed.” Runna, a popular running app that also features AI insights, generated a slightly more useful summary. It said my next run should be “easy,” one that’s perfectly timed for me to recharge. I’m sorry, but 48 hours isn’t enough time for my knees to safely heal without risk of re-opening my wounds. The in-app chatbots aren’t much better. Yesterday morning, I asked Whoop Coach if I should run today because I injured myself on my last run. It told me: “Whoop is unable to reply to the message you sent. Please try sending a different message.” I tried reframing my prompt, saying, “I’m injured and have a limp. Generate a low-intensity workout alternative while I recover.” I was prompted to contact Whoop Membership Services to continue the conversation. Oura Advisor was more helpful, noting in my daily summary: “With your Readiness dipping and recent stressors like heat, an injury, and higher glucose, your body may feel more fatigued than usual today.” It suggested I prioritize rest. When asked, “What types of movement are okay when you have an injured knee and a slight limp?” it responded with common-sense answers like a short and easy walk if there’s no pain, light stretching, and a reminder to completely rest if I feel any sharp discomfort. This is closer to an ideal response, but I had to guide it to the type of answer I wanted. The insights are so general-purpose that they benefit self-quantification newbies — and even then, only if they’re allergic to googling. My botched run is exactly the type of scenario where tech CEOs say AI insights could be most useful. In theory, I agree! It would be nice to have a competent, built-in chatbot that I could ask more nuanced questions.