Two months in the past, I uploaded greater than a decade’s value of operating knowledge into the A.I. chatbot Claude and requested for assist writing a half-marathon coaching plan. “The excellent news,” it informed me, “is your engine is big.”
That is the sort of flattery synthetic intelligence is thought for — I’m not even in my run membership’s quick group. Nonetheless, watching Claude scan the GPS recordsdata I had logged to the health app Strava and determine the highlights — a marathon private greatest, a rugged ultradistance path run — felt good.
Then issues took a flip. “The trustworthy information,” it continued, “is the place you’re ranging from proper now.” My mileage had fallen off a cliff. A standing Friday morning run with a pal was my solely constant exercise. I used to be risking harm if I attempted severe coaching too quickly.
The truth test stung, however it additionally made me assume the A.I. may be value listening to. It then requested about my targets, my present health and exercises I’d loved. Minutes later, I had a customized plan primarily based loosely on the teachings of the operating coach Jack Daniels. Quickly I used to be checking in with my A.I. coach after exercises and even adjusting my tempo mid-run to earn its approval.
A.I. has quickly flooded the health world. One business survey from final December discovered that two-thirds of gym-goers had used A.I.-powered health software program in 2025. In 2024, Strava added an A.I. exercise abstract for subscribers. The health app has additionally acquired the automated teaching program Runna, which makes use of some A.I. to change human-written plans. Final 12 months, Peloton launched an A.I. system that may rely reps and provides suggestions on type utilizing a built-in digicam.
However many individuals are merely asking general-purpose A.I. fashions to educate them towards their subsequent purpose. Whereas devoted health apps have a tendency to supply a narrower set of options, Claude or ChatGPT may be extra versatile and can not less than attempt to reply virtually any training-related query.
Private bests
For extra skilled athletes, A.I. can act as an assistant to construction and refine their very own concepts for coaching.
Daylen Yang, a 30-year-old software program engineer in San Francisco, is severe about health expertise. His private web site shows his most coronary heart fee from his final exercise, his yearly biking and operating mileage and what number of hours of sleep he acquired the earlier night time.
He first used ChatGPT for teaching final 12 months whereas getting ready for a half Ironman triathlon, which consisted of a 1.2-mile swim, a 56-mile biking time trial and 13.1-mile run. He requested the chatbot how one can enhance his private greatest by almost half-hour, and it produced a viable plan instantly, he mentioned. On race day, regardless of brutally sizzling circumstances within the Utah desert, he met his purpose.
The expertise didn’t at all times work seamlessly. When Mr. Yang got here again to ChatGPT to plan a fall marathon, the weekly mileage totals didn’t add up accurately. (A.I. typically struggles with basic math.) As soon as he ironed that out, Mr. Yang mentioned, the teaching proved its value. Past the coaching plan, the mannequin supplied steering on pacing, post-run aches and mid-race vitamin.
It additionally saved his ego in test. When he floated an aggressive purpose, the A.I. returned with one thing extra attainable. 9 weeks later, he got here inside eight seconds of that focus on.
When his mates marveled at his progress, he informed them, “It’s all because of CoachGPT,” although he credited a lot of his success to easily having a plan and executing it.
Some athletes are additionally turning to A.I teaching for energy coaching. Victoria Boyd, a weight lifter in Las Vegas, went to ChatGPT for recommendation after a knee surgical procedure had lower her most useless carry to 135 kilos from 335. She was decided to construct it again. Ms. Boyd, 44, had gotten conventional teaching prior to now, and he or she thought the A.I. plan seemed cheap: It proposed well-structured exercises that turned progressively tougher.
She saved a operating dialogue with ChatGPT to speak by means of how every session felt and to log her vitamin. She got here to understand the validation and the robust love after exhausting exercises. “It’s like: ‘You’re not at your protein purpose. It’s good to drink a shake proper now,’” she mentioned.
For Ms. Boyd, the numbers spoke for themselves — she is now as sturdy as she was earlier than her surgical procedure.
The human component
For some athletes who use A.I. to coach, expertise with human teaching has formed how they relate to the expertise.
Lengthy earlier than Chris Doenlen, 38, began working at Anthropic, the A.I. firm that makes Claude, he earned his residing as a energy coach and private coach. He now makes use of the A.I. mannequin to assist him prepare for long-distance biking within the Cascade mountain vary close to his house in northern Washington.
The coaching plan Claude made was cheap, Mr. Doenlen mentioned, and just like what a human would possibly provide you with after cautious analysis. However he was additionally conscious of what it was lacking. Good trainers depend on context and nonverbal cues, he mentioned. The A.I., then again, “is simply going off what it has from you — it exists in a pure vacuum.”
For Jon Mott, a operating coach from Lakeland, Fla., that human relationship is central to his work. However whereas coaching for a half-marathon final fall, he determined to see what he was up in opposition to. He employs six coaches who work with some 200 athletes, however he gave the job of his personal coaching to ChatGPT.
Mr. Mott, 39, is a three-time qualifier for the U.S. Olympic trials within the marathon, with a private better of two hours and 17 minutes, or roughly 5 minutes and 15 seconds per mile. The A.I. put an excessive amount of weight on that knowledge level and prescribed exercises that he “couldn’t contact,” he mentioned.
The mannequin then prompt way more pre-race relaxation than he was used to. He felt sluggish from the gun and missed his purpose by greater than 4 minutes.
Even so, he mentioned he doesn’t absolutely dismiss the expertise as a result of he sees how reasonably priced, accessible fundamental teaching might have actual worth for many individuals.
“I wish to hate on it one hundred pc,” he mentioned, “however I can’t.”
Getting over the hump
Inexperienced persons who would in any other case go with none teaching may need essentially the most to realize from A.I., Mr. Mott mentioned.
One such runner, Dustin Carl, a software program guide in Calgary, Alberta, had tried taking on the game a number of years in the past, however couldn’t make it stick. Mr. Carl, 35, mentioned the distinction this time was ChatGPT. It created a coaching plan and adjusted it in accordance with his suggestions.
Some coaches and athletes are cautious that A.I.-generated coaching plans would possibly improve the chance of overtraining or harm. However a number of of the athletes interviewed for this text discovered their A.I. to be comparatively conservative. Mr. Carl fed ChatGPT his coronary heart fee and restoration knowledge, and it dialed again the depth when his physique was careworn, which he mentioned prevented the short burnout that derailed his earlier makes an attempt.
“There’s a hump you need to recover from earlier than operating goes from depressing to one thing pleasing,” he mentioned.
The A.I. has additionally offered him with recommendation about accidents, cross-training and even emotional help. However he has additionally tried to keep up a wholesome skepticism of the expertise. “Should you’re conscious sufficient to appreciate that it’s most likely simply going to inform you what you wish to hear,” he mentioned, “you may nonetheless achieve quite a bit.”
Now he’s coaching for a full marathon. “My temper is simply so a lot better since I began operating,” he mentioned. “I feel it’s going to be a lifetime factor for me.”
