ML Coaching
A lot of people believe they can simply hand their body over to AI and get a perfect training programme. Just a few prompts, and your AI training plan for muscle growth, fat loss or marathon prep is ready.
But here's the truth: that's not how training works.
These plans are generic. They have no connection to your body, your history, or how you're feeling on a given day. They're built around averages, and in doing so, they overlook the individuality of the person.
Artificial intelligence can process data, count reps and calculate weights. But it can't feel what's happening in your body. It knows nothing about your tendons, your joints, your posture, or your sleep quality. It doesn't recognise pain, faulty loading, or fatigue.
And that's exactly the point: if you follow an AI plan, you're training blind. You don't know whether the volume is too high, whether your body can even handle the load, or whether you're working your way straight into the next overload injury.
That's exactly what this article is about. I want to show you this with clear examples, so you can see where the problems actually lie.
The difference between data and experience
As already mentioned, artificial intelligence works with data, formulas and patterns. It sees numbers, reps, heart rates, but not an actual person. And that's exactly where the problem lies.
An algorithm can tell you to train three times a week, which muscle groups to hit when, and how to increase your volume. In theory, all correct. In practice, completely impersonal. Because training isn't just made up of numbers, it's made up of feel, perception, and adaptation.
Your body changes every day. Sleep, nutrition, stress or hormones all affect how you feel and what you're capable of. AI knows nothing about that. It knows neither your history nor your injuries, it doesn't listen to your breathing, and it doesn't notice when your technique breaks down on the last rep.
An experienced trainer, on the other hand, reads your body like a book. I can see whether you're tired, whether you're able to focus well, or whether your nervous system is overloaded. I can tell when you have more in the tank and when you don't.
Data matters, no question. But it's only the foundation. Experience is what turns it into a stable system.
A real-world example: AI training plan for a marathon
And to prove this to you, I ran an experiment. A bit of context first:
I'm a personal trainer from Zurich, and I've lived for strength training for years. As a 24-time cover model, I know the highs and lows of strength sports all too well. This year, I want to take on a new challenge and run a marathon.
That's actually the reason I'm writing this article today. Because I actually started by asking ChatGPT how I should best get going:
"Create a marathon plan for beginners for me."
Two seconds later, I had a nice-looking AI training plan. Beautifully formatted, logically structured, with a clear week-by-week progression and one goal: run the marathon in 20 weeks.
Here's what it looks like:
The problems with this plan
The problem? The AI has no idea how sensitive my tendons are. The load might be fine for me, or it might cause me a lot of problems.
It also doesn't tell me that I should get proper advice first and buy the right shoes.
These might seem like small details to you right now. But trust me. If I followed this plan as a complete beginner, without engaging with the topic of marathon running in any other way, the odds are very high that I'd pick up injuries over those 12 weeks that could, in the worst case, stay with me for years.
An experienced trainer would never just send you off. He'd first check how resilient your structures are, how you move, and how you respond to load.
A proper starting point would look different:
- Screening: mobility, running technique, muscle tension
- Foundation: building mobility and stability
- Progression: slowly increasing volume and intensity
The goal isn't to exhaust you as fast as possible, it's to build your resilience for the long term.
An AI plan only recognises that you want to run.
A coach recognises whether you're actually ready for it in the first place.
The same game in strength training
The same thing happens at the gym. I'd like to give you an example here too, because it's actually even clearer in this case. (That might also be because I have more experience here, and can spot more clearly where the problems lie.)
For this, I gave ChatGPT the prompt:
"Create a training plan for muscle growth for me, five days a week."
A few seconds later, your AI training plan is ready.
Without going off on too much of a tangent, I want to show you two things about this plan that are simply not okay.
Problem #1
The first problem is the plan itself:
Specifically, the way the leg days are spread out. If you train your legs hard and intensely on Wednesday, it makes zero sense to train legs again right away on Friday. Even if it's apparently just a "short" session. Which raises the question, for me, of what difference that's even supposed to make.
It would be far more logical to do a "short" leg session on Wednesday, so you can recover enough for an intense leg session on Friday.
I also don't understand why the plan wants us to skip training on the weekend. Especially people who are busy during the week tend to shift their sessions to the weekend, simply because there's more time available then.
Problem #2
The second problem is the choice of exercises:
On paper, this all looks perfectly fine. Between 8 and 12 reps, between 3 and 4 sets, good exercises with a good effect. But!
The plan doesn't distinguish at all between a beginner and an advanced athlete. That's actually what the AI should have asked in response. How long have you been training? How much do you weigh? Do you have any joint issues?
The exercise selection is way too much volume for a beginner. If you're just starting out with strength training, it's completely enough to do just one exercise per muscle. On top of that, the plan says nothing about correct exercise execution, a warm-up, progression, or recovery. It's lacking on every front.
A lot of people follow plans like this because they look logical to someone inexperienced. But logic doesn't replace experience. Muscles only grow if your body can actually process the load, and that depends on factors AI simply can't detect: sleep, nutrition, stress, technique, posture.
A trainer pays attention to the details.
>I can see when you lose tension. I can hear when you hold your breath. And I notice when your nervous system is overwhelmed. Then we don't just "push through", we adjust.
That's how sustainable training happens:
- Technique first, weight second
- Dose the load, don't overdo it
- Strengthen weak points before they become a problem
AI can name exercises for you. But it doesn't see how you perform them.
And in the end, that's what decides between success and injury.
Why real coaching is more than just a plan
A training plan is just paper.
- It shows you what to do, but not how to do it.
- It doesn't respond to you.
- It doesn't ask why you have less energy today, or why your right shoulder hurts.
- It doesn't guide you when you're in doubt, and it doesn't stop you when you're pushing for too much.
A good coach does exactly that.
- He observes, corrects, motivates, and holds you back when it's necessary.
- He spots patterns before they turn into problems.
- He makes sure you don't just train, you understand why you're training.
Training is more than movement. It's a relationship between body and mind, and between person and coach. AI can't take responsibility off your shoulders. It can't read you, can't sense you, can't push you when you've already given up on the inside.
I don't work with people because I have to, but because I know how big the difference is when someone is truly guided. When there's someone standing beside you who never stops believing in you, even when you doubt yourself.
That's why I'll say it clearly: AI cannot replace a personal trainer. And putting too much trust in what an AI training plan can do for you really isn't good for you.
If you want real results, you have to invest, time, energy and honesty.
In and around Zurich, I work with people who want exactly that. No off-the-shelf AI training plan, no algorithm, no chance. Just training built on system, judgement and responsibility.
If you're serious, if you're truly ready to change, then don't look for a dataset, look for a coach.