How to plan a training block with ChatGPT

1.The three ways, and what each one costs

The first is what everybody does: you ask for a plan, it writes one, you copy it into wherever you train. It works once. The cost is not in that conversation — it is in every week after it, when you have to explain what already exists before it can carry on.

The second is a custom GPT: you save the instructions, the tone, the rules you want. It improves the answer and leaves the real problem untouched — it still has no access to your plan. It knows how you like a plan written; it does not know what Sofia lifted on Tuesday.

The third is connecting it as a connector. Then the assistant reads your athletes and writes into your plan: it creates the block, the weeks and the sessions. The cost is one connection, made once. That is the difference between a conversation that produces text and one that produces the plan.

2.What a chat on its own cannot know

A plan with no history is a template. The work of coaching someone is not writing week 1 — it is writing week 4 knowing what happened in the first three: which load went up, which stalled, who missed two sessions and arrives at the deload without needing one.

A chat with no access to your data has none of that. You can paste the history in, of course, but pasting a whole roster into a conversation every week is more work than writing the plan by hand. That is why almost everyone tries it once and stops: they are not giving up on the idea, they are giving up on the upkeep.

3.A five-week block, start to finish

Start with the athlete and the goal, not the exercises. "Create a 5-week hypertrophy block for Sofia, 4 days a week, upper/lower, starting next Monday" is a request a connected assistant can carry out whole — and because the block is created in one go, or fails entirely, you never end up with half of it written.

Then deal with progression. What usually works is holding the sets and adding load or a rep from week to week, with the fifth as a deload. Say so explicitly: if you do not, you will get five identical weeks, which is the worst of both worlds — it took work to build and it does not progress.

Then check it. Ask for the weekly volume per muscle group and compare it with each muscle’s target. This is where a connected assistant genuinely beats a chat: it can count what it just wrote, and tell you that back came out at eight sets when you wanted twelve.

4.The mistakes of the first conversation

The first is asking for everything in the opening message, before it has seen your data. While it does not know your roster, one request at a time is worth more: list the athletes, show the current week, and only then create. Once it has seen them, in the same conversation, you can hand it whole tasks.

The second is taking the numbers without looking. An assistant will happily write twenty-five sets of chest in a week if nobody tells it otherwise — not because it is wrong about training, but because nobody told it how much you recover. The ranges worth using are in another guide, and the arithmetic is better done by the calculator than by the conversation.

The third is not stating the constraints. How many days you actually have, what equipment is in the gym, which shoulder hurts. A plan that is perfect for a gym that is not yours is not a plan.

5.What stays yours

The decision. The assistant is fast at writing and has no opinion about your athlete: it has not seen the technique, it does not know she has slept badly since her shift changed, it will not notice the knee got worse after squats. Those are still read by you.

What it takes off your hands is the mechanical part — laying out the weeks, spreading the sessions, counting the sets, copying a template from one athlete to another. That is where the time goes, and it is not where anyone gets coached.

Tools

Guides

Frequently asked questions