Writing Prompts That Actually Work

Why Most Prompts Fail

Most people write ChatGPT prompts the way they'd type a Google search: a few words, maybe a sentence. That's understandable — it's a habit built over years of keyword-based search. But ChatGPT isn't a search engine. It's a system that responds in proportion to the quality of the instructions it receives.

A weak prompt produces a generic response. A strong prompt produces something specific, useful, and often impressive. The difference isn't length — a 20-word precise prompt routinely outperforms a 200-word rambling one. The difference is structure.

This lesson teaches you two high-leverage techniques: the CREATE framework for everyday prompts, and chain-of-thought prompting for complex tasks. Together, these two approaches will improve the quality of almost every ChatGPT interaction you have.

The CREATE Framework

CREATE is a six-part prompt structure. You don't need to use all six parts every time — but knowing them lets you diagnose why a prompt isn't working and fix it fast.

• C — Context: What's the background? Who are you, what's the situation, what does ChatGPT need to know to respond well?
• R — Request: What exactly do you want? Be specific about the task.
• E — Examples: Show what good looks like. Paste in a sample output, a style you like, or a before-and-after to calibrate the response.
• A — Action: What should ChatGPT actually produce? A list, a draft, a rewrite, a table, a plan?
• T — Tone: How should it sound? Professional, casual, warm, direct, persuasive?
• E — Expectations / Constraints: What are the limits? Word count, format, what to avoid, what to include.

You won't always need all six. A simple rewrite task might only need R + T + E (request, tone, constraints). A complex writing project might use all of them. The framework is a checklist, not a script.

Three Before-and-After Examples

The fastest way to internalize the CREATE framework is to see it in action. Here are three before-and-after pairs across common use cases.

Example 1: Email drafting

Weak prompt: "Write an email to my client about the deadline."

Strong prompt: "I'm a freelance web developer. Write a professional but warm email to a client explaining that I need to push our project deadline back by two weeks because a key third-party API has been delayed. Keep it under 150 words. Acknowledge the inconvenience, give a clear new date, and offer a 15-minute call if they have concerns."

Example 2: Content brainstorming

Weak prompt: "Give me blog post ideas about productivity."

Strong prompt: "I write a newsletter for small business owners aged 35–55 who are skeptical of tech hype but open to practical tools. Give me 10 blog post ideas about using AI to save time on administrative tasks. Each idea should have a specific, actionable angle — no vague titles like 'How AI Can Help Your Business.'"

Example 3: Data interpretation

Weak prompt: "What does this data mean?"

Strong prompt: "I'm looking at monthly website traffic data for an e-commerce store. Traffic dropped 22% in March compared to February. Here are the numbers: [paste data]. Identify the three most likely causes of this drop and suggest one diagnostic action for each. Format your response as a numbered list."

Notice what the strong prompts have in common: a clear role, a specific task, explicit constraints, and a stated output format. None of them are particularly long — they're just precise.

Chain-of-Thought Prompting

For complex tasks — analysis, planning, problem-solving, calculations — add a single instruction that transforms how ChatGPT approaches the problem:

"Think through this step by step before giving me your final answer."

Or simply: "Show your reasoning."

This technique, called chain-of-thought prompting, asks the model to reason out loud rather than jumping straight to a conclusion. It consistently produces more accurate, more reliable results for anything that requires logic, multi-step analysis, or careful judgment.

Try it on a decision you're working through:

"I'm deciding whether to hire a part-time contractor or expand my current team member's hours to handle a 20% increase in client work. Think through the pros and cons of each option step by step, considering cost, flexibility, and long-term team dynamics. Then give me your recommendation."

The reasoning that appears before the recommendation is often as valuable as the recommendation itself — it surfaces considerations you might have missed.

The Golden Rule: Clarity Beats Length

The biggest misconception about prompt engineering is that longer prompts are better. They're not. A prompt that's specific and clear in 30 words consistently outperforms one that's vague and rambling across 300 words.

When a ChatGPT response misses the mark, resist the urge to add more words to your next prompt. Instead, ask: what was unclear? Add specificity where it's missing — about the task, the audience, the format, or the tone — and nothing else.

Two habits that will permanently improve your prompt quality:

• Read your prompt before sending it. If someone other than you couldn't execute this task from your prompt, it needs more specificity.
• Iterate in conversation. ChatGPT keeps the full context of your conversation. If the first response isn't right, follow up: "Make it shorter," "Change the tone to be more direct," "Add a third option." You don't have to start over.

Good prompting is a skill you build quickly. Most people notice a significant improvement in output quality within a few days of being intentional about structure. Save the prompts that worked well — in Lesson 10, you'll build a personal prompt library so you never have to reinvent them from scratch.