01

Programming & client experience

Make me a diet plan for my client

If you want AI to help create a diet plan, you still need the right inputs and a way to deliver, track, and adapt the plan once the client starts eating real food.

By VivPublished 26 Apr 2026Last updated 26 Apr 20265 min read

02

The short answer

If you ask AI to make a diet plan for your client, the first step is not asking for meals. It is giving the system the right context: goal, bodyweight, schedule, calories, protein, preferences, constraints, and what the client actually eats now. A good AI-generated plan can save time. But the real job starts after the plan exists. Delivery, adherence, swaps, extras, and weekly adaptation. You can deliver and manage this properly through TrainedBy.

03

Use AI well

The right question is not just "make the plan."

AI gets much more useful when the coach frames the client properly. A generic 2,000-calorie request produces a generic plan. A briefed request produces something close to a real first draft. The prompt below is the shape that actually works. Paste it into Claude, ChatGPT, or Gemini, replace the bracketed values with your client's real numbers, and the output is good enough to refine in five minutes instead of forty.

04

Copy this prompt

A working AI diet-plan prompt for a coaching client.

This is the brief. Replace the bracketed inputs with your client's real numbers. The structure is what makes the output useful, because the AI builds around the constraints instead of around its averages.

You are helping me design a one-day diet plan for an online coaching client.
I am the coach. You are not the coach. I make the final decision.

Client:
- Sex: [female]
- Age: [32]
- Bodyweight: [70 kg]
- Height: [167 cm]
- Goal: [fat loss, ~0.6 kg/week]
- Training: [resistance, 4x/week, evenings ~18:30]
- Step count target: [~9,000/day]
- Allergies and refusals: [no shellfish, dislikes cottage cheese]
- Foods they like and eat now: [chicken, salmon, oats, Greek yoghurt,
  rice, sweet potato, eggs, peanut butter, berries, salads]
- Schedule: [wakes 06:30, work 09:00 to 17:00, dinner ~20:00]
- Budget: [normal UK supermarket, no specialty foods]
- Cooking: [willing, ~20 to 30 min per meal]

Targets:
- Total: ~1,900 kcal
- Protein: ~175 g
- Carbs: ~190 g
- Fat: ~60 g

Rules for the plan:
- 4 meals: breakfast, lunch, pre-training snack, dinner.
- Protein at every meal (>=25 g) except the snack.
- Training fuel in the pre-training and post-training meals.
- Use whole foods. One scoop of whey is allowed.
- Give grams for every food. Give kcal and protein per meal.
- Give a one-line note on each meal explaining the role of that meal.
- End with the day total.

Do not invent ingredients the client refused.
Do not give medical advice.
Output the plan only. No preamble.

The output you get from this prompt will look close to a real coach's first draft. It is still a draft. The coach decides what to keep, what to swap, and what to brief the client on.

05

What you'll get back

Sample output for the prompt above.

This is the kind of structure a well-briefed AI returns. Numbers are estimates and will vary slightly between models. The value is the shape of the plan, not the exact gram counts.

  1. Breakfast, around 07:30

    200 g 0% Greek yoghurt, 50 g oats, 1 tbsp peanut butter, 100 g blueberries. Roughly 470 kcal, 33 g protein. Role: protein-led start, slow carbs, holds appetite to lunch.

  2. Lunch, around 13:00

    150 g grilled chicken breast, 200 g cooked basmati rice, 150 g roasted veg, 1 tsp olive oil. Roughly 560 kcal, 45 g protein. Role: highest-volume meal, anchored away from training.

  3. Pre-training snack, around 17:00

    1 scoop whey, 1 medium banana, black coffee. Roughly 230 kcal, 28 g protein. Role: light training fuel, easy on the stomach.

  4. Dinner, around 20:00

    150 g salmon, 250 g sweet potato, 200 g salad with 1 tbsp olive oil dressing. Roughly 640 kcal, 38 g protein. Role: post-training recovery, fats for satiety overnight.

  5. Day total

    Around 1,900 kcal, 169 g protein. On target for kcal, slightly under on protein. Add a 200 ml semi-skimmed milk with breakfast, or a second yoghurt as a late snack, to land at 175 g.

06

What AI does not solve

Delivery, tracking, swaps, extras, and weekly adaptation.

AI can give you a clean draft. AI cannot watch what the client actually eats, surface the extras they did not mention, or adjust next week against reality. That is coaching work, and it lives in a system. You can deliver and manage this properly through TrainedBy. The plan goes to the client app, Snap captures real food from a photo, and the coach can adapt next week against what the client actually did instead of against what the original prompt assumed.

07

Common questions.

What should I give AI if I want a usable diet plan?

Goal, bodyweight, calories, protein, preferences, schedule, constraints, and what the client actually eats now. The better the inputs, the less fake-clean the plan becomes.

Can AI replace the coach here?

No. AI can accelerate the first draft. The coach still decides what matters, what needs adapting, and what intervention is actually right for the client.

What should happen after the plan is generated?

You should deliver and manage it properly through TrainedBy, where Snap can surface swaps, extras, and reality instead of leaving the coach guessing.

09

A good prompt saves time. A good system makes it work.

AI can help create the first version of the plan. TrainedBy is what turns that into real nutrition coaching once the client starts living their actual life.