AI and fitness: Why some athletes are using chatbots for workouts

Everyday athletes are using general-purpose chatbots and dedicated A.I. fitness tools to create and adjust training plans, and reported experiences include both measurable gains in structure and motivation and limitations such as occasional errors and lack of contextual awareness.

Everyday athletes are increasingly turning to A.I. chatbots for training advice, using both general-purpose models like Claude and ChatGPT and specialized fitness tools. The article describes a writer who uploaded years of running data to Claude and received a customized half‑marathon plan, and it notes industry developments such as Strava adding an A.I. workout summary, Runna’s acquisition, and Peloton’s A.I. rep‑counting feature. Athletes at different levels — from beginners to experienced competitors — reported that A.I. helped with planning, pacing and motivation. At the same time, several users and coaches pointed out practical limits in the technology. Key findings: - A 2025 industry survey reported that about two‑thirds of gym‑goers had used A.I.‑powered fitness software. - People use both general‑purpose chatbots for flexible, conversational coaching and dedicated apps for narrower, automated features. - Reported user successes include a software engineer who used ChatGPT to prepare for a half Ironman and meet his goal, a runner who stuck with training after getting a ChatGPT plan and is preparing for a marathon, and a lifter who rebuilt strength after knee surgery with A.I.‑generated workouts. - Coaches and experienced athletes noted shortcomings: models can make simple calculation errors, overweight past peak performances, and lack nonverbal context that human trainers provide. - Some users said A.I. behaved conservatively with intensity and adjusted plans when fed heart‑rate and recovery data, which helped prevent burnout in at least one case. Summary: A.I. tools are expanding access to structured training and daily feedback for a range of athletes while also showing limits in contextual awareness and occasional errors. Undetermined at this time.