The Difference Between Bad Prompts and Good Prompts (And Why It Matters)

The Prompt That Doesn't Work

You type "Write marketing copy." You get generic copy. Not specific to your business. The problem wasn't AI — your prompt was bad.

What Makes a Prompt Bad

Too vague: "Write marketing copy" — no product, audience, tone, length, goal. No context: "Summarize this article" — what article? Contradictory: "Professional but funny and serious and short but detailed." Assumes AI knows you: "Write the report we discussed."

What Makes a Prompt Good

Structure: 1) Role/context (who am I, what's the situation). 2) Task (what exactly to do). 3) Details (specifics). 4) Constraints (limits). 5) Output format (what you want back).

Examples: Bad → Good

Marketing: Bad: "Write marketing copy." Good: "Write a 100-word email subject and first paragraph for a SaaS targeting small business owners (2-10 employees). Product: data management tool. Goal: schedule a demo. They're skeptical. Use data/proof, not emotional appeals. Tone: professional but friendly." Data analysis: Bad: "Analyze my data." Good: Include columns, sample size, what you want to understand (e.g., which segments churn most), and that you want actionable retention insights. Code: Bad: "Write code to clean my data." Good: Specify framework (e.g., pandas), columns, specific issues (names inconsistent, email whitespace, date formats, missing values), and desired output (normalize names, trim emails, YYYY-MM-DD, flag incomplete rows, output CSV).

The Template

[ROLE/CONTEXT] I'm a [title] at [company type]. We [what you do]. [TASK] I need you to [specific task]. [DETAILS] Key details: … [CONSTRAINTS] … [OUTPUT FORMAT] I want [format]: [example].

Common Mistakes

Assuming AI remembers context — always provide full context. Being polite instead of clear — be direct. Mixing multiple requests — one request per prompt. Not specifying audience — e.g., "Explain to someone with no technical background" or "Explain to a developer."

Iterative Prompting

You don't need the perfect prompt first try. Start with "Analyze my customer data." Then refine: "Focus on why tech industry customers are churning." Then: "Which features they didn't use might have prevented churn?" Each iteration refines the output.

The Downloadable Resource

We've created a Prompt Writing Guide & Templates that includes: The prompt structure template; 20 example prompts (data, coding, writing); common mistakes and fixes; iterative refinement guide; checklist for evaluating prompts; tips by use case.

Download it here: aiforbusiness.net/resources/prompt-writing-guide

What's Next

Good prompts help you get insights. The next article, "How to Use Prompts to Extract Information from Messy Data," shows how to automate data extraction.