The SCRIPT Framework: Complete Reference for Fitness Coaches

Why Most AI Prompts Fail (and Why SCRIPT Works)

The gap between "AI-generated garbage" and "AI-generated draft I'd actually use" is almost never about the model. It's about the prompt.

When coaches ask ChatGPT or Claude to write a training program without structure, the AI has to guess almost everything — training age, goal timeline, equipment, preferred rep ranges, programming philosophy. So it defaults to the statistical middle: moderate rep ranges, common exercises, textbook periodization. Technically valid. Not particularly useful for anyone specific.

The SCRIPT framework solves this by front-loading the information the AI needs to do useful work. Instead of a vending machine ("insert prompt, receive output"), you're briefing an AI assistant the same way you'd brief a knowledgeable new coach who has never met your client. They need context before they can help.

Research consistently shows that individualization is the most important variable in training program design. A 2020 review in Sports Medicine found that individually tailored programs produce meaningfully better outcomes than standardized approaches across both strength and endurance domains. AI can't individualize without information. SCRIPT is how you give it that information.

SCRIPT Overview: The 6 Steps

Here's the full framework in one place:

The first three steps (S, C, R) go into your opening prompt. The last three (I, P, T) happen after the AI responds. Together they turn a 30-second "write me a program" into a structured briefing that gets you a usable first draft — and a workflow that gets faster every time you use it.

---

S — Specify the Role

The single highest-leverage thing you can do in a prompt is tell the AI who it is.

AI models don't have a default coaching identity. Without a role, they answer from a general fitness knowledge base — textbook-correct, but not coaching-specific. When you give the AI a role, you're activating a particular register of knowledge and communication style.

Compare:

The second prompt produces output that sounds like it came from a coach in that specialization, not a general fitness textbook. It changes exercise selection, rep range philosophy, warm-up structure, and the caveats the AI includes.

How to write an effective role

A good role specification has three parts:

• What kind of coach: Strength coach, online fitness coach, CrossFit coach, sports performance coach, rehab-informed coach, etc.
• Relevant specialization: Masters athletes, youth athletes, powerlifting, endurance, fat loss, postpartum clients — whatever narrows the expertise toward your situation.
• Coaching philosophy (optional but powerful): "Evidence-based," "RPE-driven," "periodization-focused," "functional fitness," "coach-first not program-first" — signals that significantly shape output style.

---

C — Clarify the Context

Context is client information. It's what you'd put in a client intake form — the details that make a program specific rather than generic.

The rule: you can't over-specify context. Every detail you add makes the output more relevant. Every detail you omit is something the AI has to guess, and its guesses will default to generic.

The context checklist

For programming tasks, include:

• Demographics: Age, sex, relevant health context (injuries, medications, conditions)
• Training history: How long they've been training, what they've been doing, their current numbers on key lifts
• Goal: Primary goal (lose fat, build muscle, perform at a meet, complete a race), timeline, what success looks like at the end
• Constraints: Equipment available, days per week, time per session, anything they can't do
• Preferences: Exercises they enjoy or hate, any non-negotiables in their routine
• Current training: What they've been doing most recently (avoids suggesting a massive program overhaul when continuity makes more sense)

For communication tasks (check-ins, email, social), include:

• Client personality: Analytical, motivational, needs accountability, responds to data, prefers encouragement over bluntness
• Relationship context: New client, long-term client, recent struggle or win
• Your voice: Professional, casual, direct, warm — how you actually communicate

The second context doesn't just improve program quality — it almost writes the program brief by itself. The AI only needs to fill in the structural details.

---

R — Request One Thing

This is where most coaches make the biggest mistake: they ask for too much at once.

"Write a 12-week periodized program, include nutrition guidelines, a weekly check-in template, and some social media posts I can use to attract similar clients."

That's four different prompts. When you cram them together, you get four mediocre outputs instead of one good one. AI quality degrades as scope expands in a single prompt.

The principle: one job per prompt.

This doesn't slow you down. A focused prompt takes the same time to write as a bloated one — and produces a better result that requires less revision.

Scoping correctly

For programming, don't ask for 12 weeks at once. Ask for:

• Week 1 of a 12-week program
• The periodization structure (then ask for each block separately)
• One specific block of a mesocycle

Then use subsequent prompts to continue: "Using that structure, write Week 2 with a 5% increase in total weekly volume."

For communication tasks, one prompt = one message or template. Don't ask for a check-in template, a re-engagement email, and a progress summary in the same prompt.

Format matters too

Tell the AI how you want the output delivered. This is part of the request:

• "Format the program as a table with Day / Exercise / Sets / Reps / Rest"
• "Write it as bullet points, under 150 words, in a casual tone I can send via text"
• "Give me the program in the same format as [example structure you paste in]"

Without format instructions, AI defaults to whatever structure it thinks is clearest — which often isn't what you want.

---

I — Iterate and Refine

The most important shift in mindset: the first output is always a draft.

Coaches who say "AI doesn't work" are almost always treating the first response as final. It's not. It's a starting point. The same way a first draft of anything — a program, an email, a proposal — needs revision, the AI's first output needs coaching.

Iteration is not a failure of the AI. It's the normal workflow.

How to iterate well

Be specific in your feedback. Don't say "this isn't quite right." Say:

• "Replace the Romanian deadlifts with hex bar deadlifts — my client has lower back sensitivity"
• "The volume is too high for Week 1. Drop each accessory to 2 sets instead of 3"
• "The check-in message is too formal. Make it warmer — like a text from a coach, not a form letter"
• "Add a note after each compound movement explaining the coaching cue I should be looking for"

Specific feedback = specific improvement. Vague feedback = vague revision.

How many rounds?

For most coaching tasks: 2–3 rounds is typical to get to something you'd actually use. Complex programming (full mesocycle structure, detailed periodization) might take 4–5 rounds. Simple tasks (single check-in message, exercise substitution) often need just 1.

Track your iteration count on new prompt types. If you're regularly needing 6+ rounds, your initial prompt is probably under-specified — more context up front will reduce revision loops downstream.

---

P — Personalize the Output

This is the step that separates AI-assisted coaching from AI-generated coaching.

After you've iterated the output to a point where it's structurally sound, apply your coaching eye. AI knows fitness. It doesn't know your client. It doesn't know your voice. It doesn't know the unspoken context that shapes every good coaching decision.

What personalization looks like in practice

For a program:

• Swap an exercise because you know this client always struggles with hip hinge pattern cues — choose an alternative that plays to their movement strengths
• Add your actual coaching notes in the margins: "Focus: stay tall in the hole" not generic cues
• Adjust session RPE based on what you know about their current life stress, sleep, work schedule
• Add the "why" for a deload week because this client always pushes through and needs the rationale in writing

For communication:

• Add a specific detail from their last check-in: "You mentioned your sleep has been rough — I adjusted this week's intensity accordingly"
• Match their communication register — more formal for analytical clients, more conversational for relationship-driven ones
• Remove any generic motivational language that doesn't sound like you

Personalization takes 2–5 minutes and turns a competent AI draft into something that feels like it came from their actual coach — because it did.

---

T — Track What Works

Your prompt library is a compounding asset. Every good prompt you save is time you never spend writing that prompt again.

Most coaches start by writing one-off prompts for each client. After 20 clients, they realize they're rewriting the same basic structures repeatedly with small variations. The solution is a prompt library — a collection of templates organized by use case.

How to build a prompt library

When a prompt produces an output you're happy with after 2–3 rounds:

1. Save the prompt to a Google Doc, Notion page, or your notes app
2. Label it by use case: "Strength Program - Intermediate Male - 4 Day - Hypertrophy Focus"
3. Replace client-specific details with placeholders: [CLIENT_NAME], [TRAINING_AGE], [PRIMARY_GOAL]
4. Note what worked: "Add note about knee history — Claude handles this well"

After 8–12 clients, most coaches have a library of 15–25 templates that cover 80% of their new client situations. New client programming goes from 45 minutes to 10–15 minutes. Check-in writing drops to 2–3 minutes per client per week.

What to track

Prioritize saving prompts for your most common use cases:

• New client program (by goal type: fat loss, muscle building, performance)
• Program progression week-over-week
• Client check-in response (by personality type)
• Re-engagement message (for clients who've gone quiet)
• Exercise substitution (when equipment or injury requires a swap)
• Progress summary (for clients who love data)

---

Full Prompt Examples: 3 Real Coaching Scenarios

Scenario 1: New Client Program

Scenario 2: Weekly Check-In Response

Scenario 3: Exercise Substitution

---

SCRIPT by Task Type

The framework applies to every coaching task. Here's how to adapt the emphasis for each:

---

The 5 Mistakes That Break Every SCRIPT Prompt

1. Skipping the role

The single most common omission. Without a role, the AI defaults to generic. Takes 15 seconds to add and changes everything about the output.

2. Vague context

"She's been training for a while" and "he has some injury history" are not useful. Specificity is the difference between generic advice and usable coaching output.

3. Multi-part requests

Every additional task you bolt onto a single prompt halves the quality of each part. One job per prompt, always.

4. Accepting the first output

The first output is the beginning of the conversation, not the end. If you're not iterating, you're leaving 50% of the value on the table.

5. Never saving what works

Every good prompt you don't save is one you'll write again from scratch. Your prompt library is cumulative. Start it now, even if it's just a Google Doc with three entries.

---

ChatGPT vs. Claude: Which Works Better for Each Step?

Both models work with SCRIPT. They have different strengths:

Practical recommendation: use Claude for structured program tables and check-in writing; use ChatGPT for creative exercise selection and social media. Both handle SCRIPT equally well — the framework doesn't favor either model.

---

Start Here: Your First SCRIPT Prompt in 5 Minutes

Don't try to implement the full framework on every client on day one. Start with one prompt for one task:

1. Pick one client who needs a program update or check-in response this week.
2. Open ChatGPT or Claude.
3. Write a SCRIPT prompt — Role (2 sentences) + Context (1 paragraph) + Request (1 specific deliverable + format).
4. Iterate once with specific feedback on what to change.
5. Personalize for 2 minutes before using it.
6. Save the prompt if the output was usable.

That's the whole system in one use case. From there, you're building the workflow one prompt at a time until it covers most of your coaching work.

If you want 58 pre-built SCRIPT prompts across every coaching use case — programming, check-ins, nutrition, client communication, and more — that's what the SCRIPT Toolkit is. But the framework itself is free and you can start using it right now.

Frequently Asked Questions

What does SCRIPT stand for?

SCRIPT stands for: Specify the Role, Clarify the Context, Request One Thing, Iterate and Refine, Personalize the Output, and Track What Works. It's a 6-step framework for writing AI prompts that produce useful coaching output instead of generic responses.

Does SCRIPT work with both ChatGPT and Claude?

Yes. SCRIPT works with any large language model. Claude tends to follow structured format requests more precisely. ChatGPT handles creative exercise selection well. Both benefit significantly from the SCRIPT structure — the framework isn't model-specific.

Is SCRIPT only for workout programming?

No. SCRIPT works for any coaching task: programming, check-ins, nutrition guidance drafts, progress summaries, social media content, and email communication. The framework is task-agnostic — it structures how you brief the AI regardless of what you're asking it to produce.

How long does it take to write a SCRIPT prompt?

A full SCRIPT prompt takes 3–5 minutes to write the first time. Once you have a template, filling in client-specific details takes under 2 minutes. Most coaches find total task time drops by 60–80% after they have a working prompt library.