Mr Z

Can a General AI Chatbot Build a Real Workout Plan?

Francesco Zen, founder of ZTraining · Last updated 19 June 2026

The gap isn't really about a general AI chatbot's intelligence. It's about the quality of the inputs on both sides of the conversation.

The problem starts before the AI is even involved

Most people don't have an accurate read on their own training. Someone says "I train hard" and means it sincerely, but "hard" is subjective, and a general AI chatbot has no way to check it against anything real. Tell it you want to gain muscle and get stronger, and it raises your protein target. Tell it "I'm not that overweight," and it adjusts your calories on your word alone. Every one of those inputs carries a margin of error, and a generic model has no mechanism to catch any of them.

What a generic plan gets wrong

Protein is set from bodyweight, not lean mass. Two people at the same weight with very different body fat percentages get the same protein target, which is wrong for at least one of them.
There's no body fat threshold check. Whether someone is realistically positioned to build muscle and lose fat at the same time depends on where they're starting from. A generic plan doesn't ask.
Nothing adjusts for how you actually feel that day. A real plan responds to fatigue, stress, and recovery signals. A static plan generated once doesn't.

None of this means a free AI chatbot plan has zero value. For someone doing nothing, a generic plan beats sitting on the sofa. But "better than nothing" and "built for you" are different claims, and the difference shows up within a few weeks, usually as a plateau nobody can explain.

What an embedded method changes

This is the actual difference between a general-purpose model and an AI coach built on a specific methodology.

He built Mr Z, an AI coach trained on his methodology to bring the same approach to a wider audience.