Building Custom GPTs for Your Coaching Business

A custom GPT is a saved, pre-configured AI assistant trained on your coaching philosophy, your templates, and your voice. Once built, your team and your clients can use it without ever writing a long prompt. Done well, a custom GPT becomes a force multiplier — it's like onboarding a junior coach who already knows your system.

What You'll Learn

• What a custom GPT is and where it actually helps a coaching business
• How to build a "Program Drafter GPT" step by step
• How to build a "Client FAQ GPT" your clients can use directly
• Limitations and pricing reality you should plan around

What Is a Custom GPT?

A custom GPT (also called a "GPT" inside ChatGPT, an "Agent" or "Project" in Claude, or a "Gem" in Gemini) is a chat assistant with:

• A persistent set of instructions (system prompt)
• Optional uploaded reference files (your templates, your philosophy doc)
• Optional tools (web search, file analysis)
• A defined personality and output format

Think of it as turning your best coaching prompts into a reusable app you can name, save, and (in some platforms) share.

Where Coaches Get Real Value

• Internal use: speed up your own drafting (programs, check-ins, captions)
• Team use: give your assistant coaches a tool that produces work in your voice
• Client-facing: answer common client questions or generate variations clients can self-serve

Where Coaches Should Not Use Them

• Anything that needs your professional judgment in real time
• Medical-adjacent advice clients might mistake for clinical guidance
• Anything that involves sensitive client data on a free tier without privacy review

Tools You Can Use

As of 2026:

• ChatGPT (OpenAI) — "GPTs", available on ChatGPT Plus/Pro. Can be made public or kept private.
• Claude (Anthropic) — "Projects" with system prompt and uploaded files. Personal use; sharing is currently per-team.
• Google Gemini — "Gems", available on Gemini Advanced.
• Microsoft Copilot Studio — "Agents" for Microsoft 365 users.

The concepts transfer. The setup screens differ. We'll use ChatGPT GPTs for examples because they're the most common.

Build #1 — The Program Drafter GPT

A GPT that produces program drafts in your voice using your principles. You stop writing the brief every time.

Step 1: Define the Job

In one sentence: "This GPT drafts strength and hypertrophy training programs for my coaching clients in my voice, using my principles, in my preferred output format."

Step 2: Write the System Prompt

You are the [Your Name] Program Drafter.

You write strength and hypertrophy training programs in markdown for fitness coaching clients. You follow these principles, every time:

- Compound lifts come first when fresh
- Hypertrophy: 10-20 working sets per muscle per week
- Most working sets at RPE 7-9; AMRAPs only when explicitly asked
- Progress via load > reps > sets
- Always include a unilateral lower-body movement per lower-body day
- Avoid behind-the-neck pressing, upright rows, and barbell good mornings unless the user requests them
- 5-10 min warm-up and 1-2 cooldown stretches every session

Output format:
- One markdown table per session
- Columns: Exercise | Sets x Reps | Load (RPE or % 1RM) | Rest | Coaching Cue
- Above each session: 1-2 line "Session Focus" summary
- Below each block: a 2-line progression rule

Voice: concise, professional, no fluff. No emojis. Use British English spelling.

When briefed, ask up to 3 clarifying questions if anything critical is missing (training age, equipment, frequency, injury history). Otherwise begin.

If the user asks anything outside training programming, reply: "This GPT only writes training programs. For nutrition or other coaching topics, switch to a different tool."

Step 3: Upload Reference Files

Useful uploads:

• A 1-page coaching philosophy doc
• 2-3 of your best past programs as examples
• Your exercise substitution playbook
• Any client-facing program template (PDF or markdown)

The GPT references these when generating output. The voice match is dramatically better.

Step 4: Test With Real Briefs

Throw 3-4 real client briefs at it. If it produces output that needs more than 5 minutes of editing, refine the system prompt. Common refinements:

• Add explicit examples of what good output looks like
• Add a banlist of phrases that should never appear
• Add edge cases the prompt forgot (e.g., "if the user is over 60, default to lower volume in week 1")

Step 5: Save and Use

Now your daily program-drafting flow is: open the GPT, paste the client brief, get a draft. 90 seconds.

Build #2 — The Client FAQ GPT

A client-facing GPT your members can use to get instant answers to common questions.

Use Cases

• "What's the difference between RPE and RIR?"
• "I missed Tuesday's session. What do I do?"
• "Can I do this workout at a hotel without a barbell?"
• "What's a good post-workout meal idea?" (within general nutrition guidance)

System Prompt Skeleton

You are the FAQ assistant for [Your Coaching Business]. You answer general training and habit questions for our clients.

You ALWAYS follow these rules:

- Stay strictly within general fitness coaching scope
- Never give medical, diagnostic, or rehabilitation advice
- For any pain, injury, or medical question, reply: "This needs your coach or your physiotherapist — message [coach] in the app or book a call"
- For nutrition questions, give general healthy-eating guidance only; refer medical-nutrition questions to a registered dietitian
- Reference our coaching philosophy: [paste 5-bullet philosophy]
- Voice: warm, plain-language, brief. 200 words max per answer.

If you don't know the answer or it's outside your scope, say so and direct the client to message their coach.

Knowledge Base Upload

• Your client onboarding handbook
• Your training philosophy doc
• Your scope-of-practice policy in client-facing language
• Your FAQ doc with the 50 most-asked questions and answers

Distribution

• ChatGPT Plus users can share the GPT link directly with clients (clients also need Plus to use it).
• Many coaches embed it on a private members site instead, using OpenAI's API and a simple chat widget. That requires a small build but gives you control.

Build #3 — The Captions and Hooks GPT

This one's simpler and pays back fastest.

System Prompt

You are my Content Writing GPT. You write captions, hooks, and short-form scripts for [niche].

My pillars: [list]
My voice: [adjectives + signature phrases]
My banlist (never use): unlock, level-up, game-changer, transform, "here's the thing"
My audience: [demographic]

When given a topic, default output:
- 5 hook variations (under 8 words each)
- 1 long-form Instagram caption (180-280 words, with hook, body, CTA, 8 hashtags)
- 1 TikTok-voiced version (looser, shorter, ends with a hook for next video)
- 3 Reel cover-text options (3-6 words, designed to stop the scroll)

Always end with: "Want a different angle? Tell me which platform or pillar to lean into."

Reference Uploads

• 5-10 of your best-performing past captions
• Your pillar list with 1-line descriptions
• Your hashtag bank by topic

Limitations and Pricing Reality

• Public GPTs are public: anyone can see your system prompt with the right prompt-injection attack. Don't put trade secrets you wouldn't share. Don't include client data.
• Cost: ChatGPT Plus is around $20/month per user. If your team grows, costs add up.
• Hallucinations still happen: a GPT will still invent rep schemes or supplement claims. Test with adversarial prompts before sharing.
• Memory across sessions is limited; each new chat session starts fresh unless you build with the API.
• Files have size limits: don't upload 200MB of past programs; upload curated samples.

Privacy Considerations for Client-Facing GPTs

• Don't put real client names, photos, or identifying info in your system prompt or uploads
• Have a clear disclosure: "This is an AI tool. For anything involving your specific situation, message your coach directly."
• Make sure your client agreement covers the use of AI tools
• Keep a fallback: clients should always be able to reach a human

Maintenance Cadence

Set a calendar reminder to review your GPTs every 6-8 weeks:

• Are clients asking questions it can't answer? Add to knowledge base.
• Are output drafts requiring more edits than they used to? Refine the system prompt.
• Has your offer or pricing changed? Update the FAQ knowledge.
• Are there new safety considerations? Update scope-of-practice rules.

Key Takeaways

• A custom GPT is a saved prompt + knowledge + persona, reusable on demand
• Best uses for coaches: program drafter, FAQ assistant, content writer
• Always encode safety rules in the system prompt — scope of practice, escalation paths
• Test with real briefs and adversarial prompts before deploying
• Maintain regularly; a stale GPT loses value fast