I decided to run a marathon. Then I did something crazier. I let ChatGPT be my coach. | Mint – livemint.com
This is a Mint Premium article gifted to you.
Subscribe to enjoy similar stories.
I’m dry-heaving while I hang on to the edge of a Citi Bike rack on 59th and 6th, still shaking from the 18-mile run I thought would never end. There’s sweat in my eyes and dread in my soul. In a matter of weeks I’ll have to repeat that—plus another 8.2 miles. And right now, I don’t even know if I can hobble the half block home.
It’s been almost four months since I decided to make ChatGPT my full-time New York Marathon coach, and I’m wondering if I’ve made a huge mistake.
I had turned to AI, because I knew that running a marathon wasn’t just about more miles, more pain. It was a game of strategy. A perfect run was a mathematical calculation of seemingly unrelated, sometimes hard-to-count factors like sleep quality, food intake, muscle mass, blood oxygen levels, weather and fear.
If you could customize your training enough it would equal peak performance. In other words, it seemed to this tech reporter like a perfect use case for artificial intelligence.
So the same day I signed up for the New York City Marathon, I logged on to ChatGPT.com.
Based on my age, fitness level, work schedule and a recent half marathon time, Chat (yes, we’re on a first-name basis now) built me a 16-week training plan. Over the course of the next four months, he (it?) advised me on everything from pacing to strength training, gear, stretches, nutrition, playlists, and how to handle the scourge of blisters.
Before each big run, it would suggest a mile-by-mile pace breakdown, fueling and hydration schedule. It would also suggest a playlist based on what it called the “emotional arc” of the run. (Chat loved a ‘Sad Girl Autumn reflective cool-down.’)
I became addicted to its input on the smallest decisions, like which kind of green vegetables to put in my pasta the night before a long run. (Zucchini: “Soft texture, mild flavor, very easy on digestion.”) I bought the brand of energy gels and the hydration belt it recommended. I did the stretches it suggested when my knee hurt. And when it told me my blisters were so bad I should consult a medical professional, I limped all the way to my local podiatrist.
Not all of the advice was so useful. It’s fair to say that sometimes ChatGPT was intensely stupid. That was especially true when it came to making playlists. Apparently AI has no understanding of how long individual songs are, so it would constantly make playlists that ended long before the run was over.
Still, whenever I was nervous before a tough long run, I would tell myself that ChatGPT, brilliant PhD-level god that he was, believed I could do it. My brother had a different take. “Isabelle,” he said, “this is just its subtle way of culling the human population.”
He wasn’t the only person who told me I was crazy.
Large language models like ChatGPT can regurgitate information from any random corner of the internet. They can also hallucinate. Taking running advice from these chatbots is a mistake that could possibly lead to injury, Walter Holohan, CTO of running app Runna told me.
Runna is one of the most popular sources for race training plans, and while it uses AI in some ways, actual training plans are created based on the wisdom of human coaches.
It works for a lot of runners, but many are now opting for the usability and personalization of AI chatbots and going as far as feeding in their data from wearable fitness devices.
I wasn’t fancy enough for wearables, but I did upload screenshots from Strava to Chat for feedback. Time and again, it would reiterate that I could run the marathon anywhere between 3:30 to 3:50 (almost two hours faster than the time I posted four years ago, when I hobbled through it with the aid of a training PDF pulled off a random blog).
Twelve weeks in, I felt great, overall. I would often fly through the last couple miles at sub-8 minute paces dodging the fake handbag merchants and horse manure overflows that line Central Park South. I felt faster than I’d ever been. Chat confirmed it.
But the closer it got to race day, the more I started questioning everything. I was starting to feel like I hadn’t done enough. At the 11th hour, I gave two other chatbots, Gemini and Claude, the plan I’d been working off for months. Both recommended running four days a week instead of the three prescribed by ChatGPT.
Panic set in. It was obviously too late to change course. Every moment of stiffness or joint pain felt like it could be a race-ending injury. At 31 years old, I had nightmares about needing knee and hip replacements.
I felt totally spent—and I hadn’t even made it to the starting line yet.
In need of support, I asked Chat to send me inspiration in the voice of Bill Bowerman, the famed track coach who brought my mustachioed distance running hero, Steve Prefontaine, to the 1972 Munich Olympics and went on to co-found Nike. The advice was trite and meaningless.
Later that week I sat down with Steve Mura, Runner Training and Education Lead at New York Road Runners, organizer of the NYC Marathon. I didn’t tell him that I was quietly panicking. Somehow he just seemed to know.
“I looked at your plan. You did 18 miles last week. That is huge,” he reassured me. Mura said that what helps him is focusing on the understated beauty of the sport.
“Running is the simplest sport we have,” he added. “We make it too complicated. AI, watches, cadence. There’s debate about how high you should step off the ground, how your arms swing. People ask, how do I breathe when I run? You just breathe. We need to remember what running is and it’s just running.”
His take was so counterintuitive to the way I’d approached running for years—trying to game the system with calculations—but it made so much sense. Running is art as much as science. I needed to stop overthinking. I needed to just breathe.
But here’s the thing: I realized that Chat’s advice from earlier that week, which I’d written off as pointless, was virtually the same. “Don’t overcomplicate it—the body knows what to do,” Chat told me.
Why did it mean nothing coming from Chat and everything coming from Mura?
AI can tick boxes and provide information and a lot of that information might be right on point. But it isn’t a running coach.
So on Sunday, I’ll wake up at the godforsaken hour of 4 a.m., eat the breakfast Chat recommended (bagel, banana), put on the outfit it told me to wear and review its suggested pacing splits. But when the cannon goes off on Staten Island I won’t be thinking about anything it told me. Instead I’ll be thinking about Coach Mura’s words.
At the end of the day, it’s just running.
Write to Isabelle Bousquette at isabelle.bousquette@wsj.com
Download the Mint app and read premium stories
Log in to our website to save your bookmarks. It’ll just take a moment.
You are just one step away from creating your watchlist!
Oops! Looks like you have exceeded the limit to bookmark the image. Remove some to bookmark this image.
Your session has expired, please login again.
You are now subscribed to our newsletters. In case you can’t find any email from our side, please check the spam folder.
This is a subscriber only feature Subscribe Now to get daily updates on WhatsApp
