Mega-Prompts: How to Build Multi-Step AI Instructions

What Is a Mega-Prompt?

A mega-prompt is a structured, multi-section prompt that produces comprehensive output in a single generation. Instead of going back and forth with the AI, you provide all context, requirements, and formatting instructions upfront.

Think of it as a detailed creative brief — the more precise your input, the better the output.

Why Mega-Prompts Work

Standard prompts produce standard output. Mega-prompts work because they:

1. Eliminate ambiguity — every section has specific requirements
2. Set quality expectations — you define what "good" looks like
3. Reduce iterations — get 80-90% right on the first try
4. Enable complex outputs — business plans, research reports, full strategies

The Anatomy of a Mega-Prompt

Every effective mega-prompt has these components:

1. Role & Context

You are a [specific role] with [years] of experience in [domain].
Your task is to create a [output type] for [audience].

2. Input Variables (what the user provides)

## Context
- **Company**: [name]
- **Industry**: [sector]
- **Target audience**: [who]
- **Budget**: [range]

3. Output Sections (what the AI generates)

## Generate the Following Sections:

### Section 1: Executive Summary
- Length: 200-300 words
- Include: key findings, recommendation, timeline
- Tone: professional, data-driven

### Section 2: Market Analysis
[specific requirements for this section]

4. Formatting Instructions

## Format Requirements:
- Use markdown headers for each section
- Include tables where data comparison is needed
- Bullet points for lists of 3+ items
- Bold key terms and metrics

5. Quality Constraints

## Rules:
- Every claim must include a supporting data point
- Use specific numbers, not vague language ("15% increase" not "significant growth")
- If information is assumed, label it clearly as an assumption

Building Your First Mega-Prompt

Step 1: Define the Output

Start with the end. What exactly do you want to receive? Write out the table of contents of your ideal output.

Step 2: List the Inputs

What information does the AI need from you? Create fill-in-the-blank variables for everything context-dependent.

Step 3: Write Section Specs

For each section of your output, specify:

• What it should contain
• How long it should be
• What format to use
• What quality bar to meet

Step 4: Add Guardrails

Rules that prevent common AI failure modes:

• "Don't use generic statements"
• "Include specific examples from the industry"
• "If you don't have data, say so — don't make it up"

Example: Converting a Basic Prompt to a Mega-Prompt

Basic prompt:

Write a marketing plan for my SaaS product.

Mega-prompt version:

You are a CMO-level marketing strategist with 15 years of B2B SaaS experience.

## Context
- Product: [name and one-line description]
- Stage: [pre-launch / just launched / growing]
- Monthly budget: [$X]
- Current MRR: [$X]
- Target customer: [ICP description]

## Generate a 90-Day Marketing Plan:

### 1. Positioning Statement (one paragraph)
Format: For [target], [product] is the [category] that [key benefit] unlike [competitor], we [differentiator].

### 2. Channel Strategy (table format)
| Channel | Budget % | Expected CAC | Timeline to Results |
Evaluate: SEO, Paid Search, LinkedIn, Content, Community, Partnerships

### 3. Content Calendar (Month 1)
| Week | Topic | Format | Channel | Goal |
4 weeks of specific content pieces.

### 4. KPIs & Milestones
Month 1/2/3 targets for: MQLs, SQLs, Demo requests, New customers, CAC

### 5. Quick Wins (Week 1)
3 things to do THIS WEEK that cost under $100.

## Rules:
- Budget allocations must sum to 100%
- Every recommendation needs estimated ROI
- Include at least one unconventional channel

Common Patterns in Great Mega-Prompts

Pattern	What It Does	Example
Table format	Forces structured comparisons	"Compare in table format: Feature, Priority, Effort, Impact"
Numbered sections	Ensures completeness	"Generate sections 1-8, each with..."
Word count targets	Controls depth	"Section 1: 200-300 words"
Examples of good/bad	Sets quality bar	"Good: specific, data-driven. Bad: generic, vague"
Checklists	Verification	"Ensure each section includes: data, example, actionable takeaway"
Variables	Reusability	"Replace [COMPANY] and [INDUSTRY] to customize"

Tips for Better Mega-Prompts

1. Test with different inputs — a great mega-prompt works for any valid input, not just one specific case
2. Iterate the prompt, not the output — if the output is wrong, fix the prompt
3. Use the output as a starting point — mega-prompts get you to 80%, then you refine
4. Save and version your prompts — they're intellectual property
5. Share on Promptsy — the community can improve them further