Imagine having an AI as your personal fitness coach – sounds futuristic, right? Well, I decided to put ChatGPT to the test and see if it could truly replace a human trainer. But here’s where it gets controversial: while it delivered a detailed training plan, the results were far from perfect. Let’s dive into what happened and why this might not be the fitness revolution we’re hoping for – yet.
If you’ve been online in the past few years, you’ve likely encountered ChatGPT’s infamous prompt: ‘Where should we begin?’ Launched in November 2022, OpenAI’s chatbot sparked a race in generative AI, with tech giants like Microsoft, Google, and Meta quickly following suit. At first glance, it seems like a fancy search bar or a tool for drafting formal emails. But with the right prompts, it promises to unlock capabilities far beyond that – including, apparently, coaching.
The fitness world, in particular, is buzzing with AI integration. Apps like Strava’s Runna use AI to monitor your training progress and suggest adjustments, while HumanGo AI offers a chatbot named Hugo to create and tweak workout plans. And this is the part most people miss: these tools are positioning themselves as alternatives to human coaches. But how effective are they really? I decided to find out by asking ChatGPT to design a marathon training plan for me.
As someone who’s run marathons with a personal best of 2 hours and 43 minutes, I’m no stranger to structured training. I’ve used everything from free online guides to personalized plans crafted by a coach for events like the Paris-Roubaix Challenge and Manchester Marathon. The difference between a generic plan and a tailored one is night and day – the latter offers guidance, accountability, and adaptability that a one-size-fits-all approach simply can’t match.
To test ChatGPT, I asked for a 16-week sub-4-hour marathon training plan, specifying I run 3-4 times a week. Its response was impressively detailed: goals, weekly breakdowns, the full 16-week plan, explanations of key workouts, and tips on warming up, cooling down, hydration, and fueling. Each week included four runs – three easy and one hard, with a long Sunday run building up to 18 miles. Recovery weeks were also factored in. So far, so good.
But here’s the catch: upon closer inspection, the plan felt basic. While it followed a periodized, polarized approach, it only incorporated two intensities – tempo and 5K pace – for the entire 16 weeks. Crucial elements like anaerobic work (sprints, HIIT) and strength training were missing, despite their proven benefits for endurance, body composition, and running economy. Bold statement alert: this plan would likely bore experienced runners and fall short for anyone seeking significant improvement.
Faye Johnson, a running coach and Level 4 PT at Maximum Mileage Coaching, echoed my concerns. ‘It’s very generic and repetitive,’ she said. ‘I’d include a variety of sessions and gradually introduce different types of speed work – progressive runs, maximal intervals, threshold intervals – to keep things engaging.’ She also pointed out the absence of hill workouts, which are essential for building strength and speed, especially for undulating marathon routes.
That said, Johnson admitted the plan wasn’t terrible. Its structure and weekly breakdown were solid, and for someone comfortable with their pacing, it could work. ChatGPT’s ability to generate such a plan in seconds is impressive, likely drawing from countless publicly available training resources. When asked, it claimed to base the plan on ‘reputable coaching sources and methodologies,’ including input from RRCA and USATF-certified coaches.
But here’s the controversial part: while AI can provide structure, it lacks the human touch that makes coaching effective. Accountability, for instance, is a major factor. With a coach, you know they’ll see if you’ve skipped a session – a motivation that’s missing with AI. Plus, a human coach can spot nuances in your training, like whether your ‘easy run’ is actually too intense, and adjust accordingly. AI, on the other hand, relies on rigid pacing guidelines that don’t account for variables like weather or fatigue.
This rigidity can lead to overtraining and injury. Johnson noted, ‘Some runners don’t know when to back off, and a coach is there to tell them.’ When I simulated a sore calf injury with ChatGPT, it offered sensible advice – replace speed work with easy runs, avoid running through pain, and seek physio if needed. But it couldn’t diagnose the root cause, meaning the issue could resurface later.
The biggest flaw? ChatGPT relies heavily on the user. You need to craft precise prompts, understand the plan’s intent, and know when to adapt it. Even then, trusting a free AI tool that’s only been around for three years is a leap of faith. As Johnson put it, ‘How can AI account for all the nuances of a personalized activity plan?’
For now, I’ll stick to using ChatGPT for life-admin tasks. Until it can replicate the insight, feedback, and credentials of a real coach, it’s not quite ready to take the reins. But here’s the question I leave you with: Do you think AI will ever fully replace human coaches, or will it always fall short in areas that require empathy, intuition, and experience? Let’s discuss in the comments!
