Prompt Framework
ChatGPT Prompt Framework: Structure Every Prompt for Better Output
The difference between a generic ChatGPT response and a genuinely useful one almost always comes down to prompt structure. This framework gives you a repeatable system for writing prompts that produce better first drafts, require less editing, and handle more complex tasks reliably.
The core framework: Role · Task · Context · Audience · Format · Constraints
Every strong ChatGPT prompt can be built from six elements. You don't need all six in every prompt -- but the more of them you include for complex tasks, the better the output tends to be.
Role
What it is: Telling ChatGPT what kind of expert it should act as. This shapes the vocabulary, assumptions, and depth of the response.
Why it matters: ChatGPT is a generalist. Giving it a specific role focuses its response on the knowledge and communication style most relevant to your task. "Act as a senior UX researcher" produces very different output than "help me with this."
Examples: "Act as an SEO strategist," "Act as a professional resume writer," "Act as a direct-response copywriter with 10 years of B2B experience."
Task
What it is: The specific job you want done -- stated precisely, with a clear deliverable.
Why it matters: Vague tasks produce vague output. "Help me with marketing" is unusable. "Write 5 Facebook ad hooks for a dog training course targeting first-time dog owners" is specific enough to produce something you can actually use or refine.
Context
What it is: The background information ChatGPT needs to do the task well -- your industry, situation, what you've already tried, or what you're working with.
Why it matters: ChatGPT doesn't know your business, your customers, or your constraints. The context you provide replaces the assumptions it would otherwise make -- which are rarely as accurate as your actual situation.
Audience
What it is: Who the output is for -- their expertise level, role, concerns, and expectations.
Why it matters: A technical explanation for an engineer and an explanation of the same topic for a CFO require completely different vocabulary, depth, and framing. Specifying the audience eliminates this guessing.
Format
What it is: How you want the output structured -- bullet points, numbered steps, paragraphs, a table, JSON, a polished draft, or a specific length.
Why it matters: Without format guidance, ChatGPT defaults to paragraph prose, which is often not the most useful format for the task. Specifying format makes output immediately usable.
Constraints
What it is: Rules, limits, or things to avoid -- word counts, style restrictions, what not to include, or quality requirements.
Why it matters: Constraints prevent the most common failure modes -- responses that are too long, that include what you specifically don't want, or that take the easy generic answer instead of the specific useful one.
The framework formula
Act as [role].
Task: [specific deliverable].
Context: [relevant background].
Audience: [who this is for].
Format: [how to structure the output].
Constraints: [what to avoid or limit].Framework applied: 6 real examples
Business strategy
Act as a practical small business advisor. Task: create a focused 90-day growth plan for a local painting contractor in Sacramento. Context: annual revenue ~$180K, 2 employees, primary challenge is inconsistent lead flow, currently relying on word-of-mouth. Audience: the owner who will execute this alone. Format: 3 monthly priorities, 3 specific actions each, 1 tracking metric per month. Constraints: no paid ads over $200/month, no strategies that require hiring.
Content writing
Act as a direct-response copywriter. Task: write a 150-word above-the-fold section for a landing page offering a free SEO audit. Context: targeting small business owners who know they need SEO but don't know where to start. Audience: non-technical local business owners. Format: H1 headline, H2 subheadline, 3 benefit bullets, CTA button text. Constraints: no jargon, no false urgency, no generic claims like 'boost your rankings.'
Research and analysis
Act as a market research analyst. Task: summarize the key decision factors for someone buying premium dog food for the first time. Context: I'm writing copy for an e-commerce store selling a $65/month subscription dog food brand. Audience: first-time premium pet food buyers, household income $75K+. Format: organized by emotional and rational factors, with a short note on the most common buying objection. Constraints: base analysis on observable patterns, not invented statistics.
Email drafting
Act as a professional business writer. Task: write a follow-up email to a client who received a proposal 6 days ago and hasn't responded. Context: the proposal was for a $12,000 website redesign, the initial call went well. Audience: a marketing director at a 50-person company. Format: subject line, 80-word body email. Constraints: no desperation, no 'just checking in,' no passive-aggressive nudge -- curious and confident only.
Code explanation
Act as a senior software engineer explaining to a junior colleague. Task: explain what this Python function does and identify any edge cases I should handle. Context: [paste function]. Audience: a junior developer with 1 year of Python experience. Format: (a) plain-language explanation of what it does, (b) 3 edge cases not currently handled, (c) one suggested improvement. Constraints: no assuming they know advanced Python concepts.
SEO content brief
Act as an SEO content strategist. Task: create a content brief for an article targeting the keyword 'local SEO for plumbers.' Context: the target site is a plumbing company in Phoenix, DA ~25, 15 existing pages. Audience: a freelance writer with SEO knowledge. Format: title tag, meta description, H1, search intent analysis, 6 H2 sections with content direction, word count, and 3 internal linking opportunities. Constraints: each H2 must address a specific reader question, not just a topic label.
When to use the full framework vs. a shorter prompt
Not every ChatGPT prompt needs all six elements. A quick brainstorm or a simple factual question works fine with a minimal prompt. Apply the full framework when:
- The task is complex or multi-part
- The output needs to be usable without significant editing
- The audience is specific (not a general reader)
- You've tried a simpler prompt and the output was too generic
- The stakes are higher -- client work, professional content, or important decisions
Iterative prompting: the second prompt matters as much as the first
The framework produces a better first draft -- not always a final one. The most effective ChatGPT users treat the first response as a foundation and then refine with specific follow-up instructions:
- Add specificity: “The third bullet is too vague -- add a concrete example from the construction industry.”
- Adjust tone: “Rewrite this to sound like a founder talking to a peer, not a consultant presenting to a board.”
- Change format: “Convert this into a table comparing the three options side by side.”
- Self-critique: “Review this response and identify the two weakest parts. Then rewrite those sections.”
Common framework mistakes
- Vague roles. 'Act as an expert' gives less direction than 'Act as a B2B SaaS marketing manager with experience in content-led growth.' Specificity in the role produces more relevant output.
- Missing context. ChatGPT doesn't know your product, your customers, or your specific situation. The context block is where you fix that. Leaving it empty forces the model to guess.
- No format specification. ChatGPT defaults to paragraphs. For most business tasks, bullet points, numbered steps, or structured sections are more immediately useful. Always specify.
- Asking for too much in one prompt. A prompt that asks for a full marketing strategy, landing page copy, and 10 blog titles in one go produces shallow output for each. Break complex requests into sequential prompts.
- Accepting the first response as final. The framework produces a strong first draft, not a final product. Always read, evaluate, and give specific refinement instructions.
Pro Mode: adding quality bars and refinement instructions
For high-stakes prompts, add a quality bar that tells ChatGPT what "done well" looks like for this specific task:
Quality bar: This response will be used directly as client-facing copy.
It must be: specific enough that it couldn't apply to a different company,
free of marketing clichés, under 150 words, and immediately publishable
without editing.Adding a quality bar shifts the model from "complete the task" to "complete the task to a standard." The difference in output quality is consistent and significant.