Anatomy of a Great Prompt

Every effective prompt contains some combination of these five components:

1. Role — Who should the AI be? "You are an experienced marketing strategist..."
2. Context — What background information does it need? "I'm launching a SaaS product for small businesses..."
3. Instruction — What exactly should it do? "Create a 30-day launch plan..."
4. Format — How should the output look? "Present as a table with columns for Week, Task, Channel, and KPI"
5. Constraints — What should it avoid or prioritize? "Budget is $5,000. Focus on organic channels. No paid ads."

You don't need all five for every prompt, but the more you include, the better the output.

Bad prompt: "Write me a blog post about AI"
Problem: No role, no context, no format, no constraints. The AI has to guess everything.

Good prompt: "You are a tech blogger writing for a non-technical audience. Write a 500-word blog post explaining how small business owners can use ChatGPT to save 5 hours per week. Use a conversational tone, include 3 specific examples, and end with a clear call-to-action. Avoid jargon."

The good prompt specifies the role (tech blogger), context (non-technical audience, small business), instruction (500-word blog post), format (conversational, 3 examples, CTA), and constraints (no jargon).

The #1 rule of prompt engineering: be specific.

Vague prompts get vague responses. Specific prompts get specific, useful responses.

- Instead of "make it better" → "improve the opening paragraph by adding a hook question and a surprising statistic"
- Instead of "write code" → "write a Python function that takes a CSV file path and returns the top 5 rows by revenue, handling missing values gracefully"
- Instead of "summarize this" → "summarize this article in 3 bullet points, each under 20 words, focusing on the implications for software developers"