Fitness, artificial intelligence, and the role of chatbots in training
Artificial intelligence is racing into the fitness world, and it is no exaggeration to say that a lot of people already have a virtual coach in their pocket. Some call it CoachGPT, others use similar names, but the logic is the same: you talk to a chatbot as if it were a personal trainer, ask for help building training plans, interpreting performance data, adjusting training load and, sometimes, even hearing some hard truths about a lack of consistency.
In the original New York Times article, the starting point is very clear: everyday athletes, not just professionals, are using AI models like ChatGPT and Claude to hit real goals in running, triathlon, strength training, and cycling. Instead of relying only on traditional apps or generic programs, these people started uploading years of running data, like Strava history, average pace, past marathons, heart rate, sleep, and even post-workout feelings, asking the AI to create a personalized plan.
In many cases, this AI does not show up as a polished app full of pretty charts, but as general-purpose models such as Claude or ChatGPT that the user customizes to act like a personal coach. They read GPS files, identify performance peaks, volume drops, injury risks, and return a structured summary with an action plan. Interestingly, even with its reputation for being overly flattering, the AI also delivers some tough feedback when it notices training volume has crashed or when the athlete wants to progress too fast.
AI and fitness: from dedicated apps to general-purpose models
The original article points out that AI has already become standard in several fitness platforms. An industry survey cited by the paper showed that, by 2025, about two-thirds of gym-goers had already used some kind of AI-based training software. That covers everything from apps that suggest automatic workouts to more advanced systems that analyze patterns of effort, recovery, and performance.
Some concrete examples in the story:
- Strava added an AI-powered workout summary for subscribers, automatically interpreting what happened in the session and explaining it in simple language.
- Strava also acquired Runna, an automated training program that mixes human-written plans with AI-based adjustments, adapting load as the athlete responds.
- Peloton launched an AI system capable of counting reps and giving form feedback using the equipment’s built-in camera, bringing the experience closer to what a live instructor would do.
But the core of the report is not these ready-made apps, it is the turning point that happens when people start using general-purpose AI models as virtual coaches. Instead of a fixed feature list, these models try to answer any question about training, running, strength work, triathlon, basic nutrition, and race strategy, even when the topic is complex. That flexibility makes them a kind of Swiss army knife for amateur athletes who want more context and explanation without being locked into a rigid interface.
From data upload to a half marathon plan
One of the cases described involves a runner who dumped more than ten years of running data into Claude and asked for a half marathon plan. The AI analyzed Strava files, saw a strong background with a solid marathon and even a trail ultramarathon, but also identified that his recent training volume had crashed. Instead of just handing out empty praise, the model did a reality check: it acknowledged that his engine was good but warned that coming back too hard could raise injury risk.
From there, the chatbot asked about goals, current routine, and the types of workouts the athlete usually enjoys. In a few minutes, it built a plan loosely based on the principles of a well-known running coach, Jack Daniels, famous for his structured methods. The most interesting part is that the relationship did not stop there: the athlete started “reporting” workouts to the AI, adjusting pace mid-run to match suggested parameters, and chatting with the model to revise the plan as his body responded.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
AI as a high-level assistant: triathlon, marathons, and strength training
Once an athlete has some experience, AI stops being just a motivational tool and turns into a technical assistant, helping organize ideas and fine-tune details that make a difference on race day.
Planning an Ironman with chatbot help
The article brings up the example of Daylen Yang, a 30-year-old software engineer obsessed with tech and data. On his personal site, you can find numbers like heart rate from his last workout, annual running and cycling mileage, and even sleep hours. He turned to ChatGPT to prepare for a half Ironman, including swim, bike, and run.
His goal was ambitious: cut his total time by roughly 30 minutes. He asked the chatbot for a detailed plan to hit that mark. The model instantly returned a training cycle structure that he considered plausible. On race day, even under brutal desert heat in Utah, he hit his target.
Of course, not everything was perfect. When he later asked for a marathon plan for the fall, he noticed the model messed up simple weekly mileage sums, which fits what we already know about LLMs: they can struggle with basic math. After fixing those inconsistencies, he stuck with the plan, got guidance on pacing, post-run soreness, and race nutrition, and finished just seconds off his goal, in a very competitive result.
Rebuilding strength after surgery with AI support
Another striking case is Victoria Boyd, a lifter from Las Vegas. After knee surgery, her deadlift dropped from 335 to 135 pounds. Determined to regain strength, she had previously worked with human trainers and therefore had a good sense of what makes sense in terms of periodization.
When she asked ChatGPT for a plan, she carefully checked whether the proposal followed one of the most important principles of strength training: progressive overload, gradually increasing difficulty. According to the account, the plan looked coherent, with a well-organized training structure, solid progression, and a balanced distribution of volume and intensity.
She kept an ongoing dialogue with the chatbot, logging how she felt, how her knee was doing, whether there was unusual pain or just normal discomfort from heavy lifting, and using AI to monitor her diet. Several times, the model recommended more protein intake and even suggested a shake when it noticed she was missing daily targets. The result, according to Victoria, was clear: she returned to her pre-surgery strength level.
What AI still cannot see: the importance of the human factor
Not everyone in the article sees AI as a replacement for a traditional coach. On the contrary, several interviewees emphasize that even if the training generated by the machine is technically acceptable, it lacks something only a human can offer.
A coach who became an AI company employee
Chris Doenlen, now an employee at Anthropic, the company behind Claude, previously worked as a strength coach and personal trainer. He uses the company’s own AI to prepare for long cycling challenges in the mountains near his home in northern Washington.
According to him, the AI plan is reasonable and resembles what an experienced human coach might put together after studying the case. However, Chris points out that a good human coach also reads context and nonverbal signals: posture, facial expression, body language, tone of voice, excitement, or dips in motivation. The AI does not see any of that. It operates in a data vacuum, reacting only to what the athlete writes and the numbers it receives.
When AI overestimates an experienced athlete
The piece also highlights the case of Jon Mott, a running coach in Florida and elite athlete with three appearances at US Olympic Marathon Trials. He intentionally decided to test the AI: even though he has six human coaches on his staff and around 200 athletes under guidance, he let ChatGPT handle his own half marathon plan, as if he were any regular client.
The model leaned heavily on his best marathon time, 2:17, which is about 5:14 per mile. Based on that, it suggested workouts at such aggressive paces that Jon himself described them as impossible to execute. On top of that, before the race, the chatbot recommended a longer taper than he was used to.
The result: he got to the starting line feeling heavy and sluggish and finished the half marathon more than four minutes slower than his target. Even so, Jon does not dismiss AI completely. He sees the technology’s potential to provide a basic level of guidance for lots of people who otherwise would have no access to any coaching at all.
Where AI can help the most: beginners and consistency
Among all the profiles in the article, one of the groups that benefits the most from AI are beginners who have tried to train on their own, failed, and now find a middle ground between giving up and paying for expensive one-on-one coaching.
Getting over the running hump with digital support
One example is Dustin Carl, a software consultant in Canada. He had tried getting into running a few years earlier but could not sustain the habit. On his next attempt, he turned to ChatGPT to create a plan and then tweak it as he sent in feedback.
A common concern among coaches is the risk that AI-generated plans might push people too hard, with overly high volume, uncontrolled intensity, and injuries as a result. Interestingly, what showed up in this runner’s account was the opposite: by feeding the chatbot with heart rate and recovery data, he noticed the AI tended to be conservative, dialing back intensity during phases when his body showed more accumulated stress.
He mentions an idea every runner knows well: there is a psychological and physical wall in the first weeks or months, when everything feels uncomfortable, slow, and heavy. If you get past that, you start to enjoy the routine. In his case, AI helped regulate effort, prevent early burnout, and keep him going until running started to feel good.
Beyond the physical side, this kind of chatbot also offered guidance on common aches and pains, cross-training ideas, stretching suggestions, and even some emotional support, reinforcing more cautious decisions. Dustin himself, however, makes an important caveat: he says you have to remember that AI tends to tell you what you want to hear, and keeping a critical mindset helps you get the best from the tool without falling for illusions.
Balancing technology, body, and experience
In the end, the landscape at the intersection of fitness and artificial intelligence is less about replacing coaches and more about expanding what is possible. Models like Claude and ChatGPT are already able to:
- read large volumes of training data and organize them in simple language;
- create coherent plans for many cases, especially beginners and intermediates;
- help experienced athletes structure training cycles and test ideas;
- explain concepts around training, basic nutrition, and recovery in an accessible way;
- act as an on-call pocket coach when there is no human support available.
At the same time, the stories show that AI:
- can misjudge training load and pacing, especially when it relies too heavily on a single data point, like a personal best;
- cannot see nonverbal cues, emotion, unusual pain, or personal life context;
- depends entirely on input quality: if the data is bad or exaggerated, the guidance will be off;
- risks encouraging overdependence if athletes stop listening to their own bodies.
The original article leaves a message that applies both to beginners and to athletes with plenty of long-distance experience: artificial intelligence can be a powerful ally for building consistency, organizing training, and making sense of data, but it still works best when it walks side by side with critical thinking, self-awareness, and, when possible, human professional support.
At the end of the day, it is still a flesh-and-blood body crossing the finish line. AI can help pick the route, dial in the pace, and avoid some common mistakes, but daily discipline, reading your own signals, and balancing effort with recovery are what really determine how far each athlete can go.
