You’ve probably already noticed: garbage in, garbage out.
Ask ChatGPT a vague question, get a vague answer. Ask it well, get something actually useful.
That skill is called prompt engineering, and it’s the difference between AI being helpful and AI being a waste of time.
## What Good Prompts Look Like
**Bad prompt:**
“Write me something about AI”
**Good prompt:**
“Write a 200-word overview of how large language models work, aimed at someone who has no technical background. Use analogies instead of jargon. Include one practical use case for small business owners.”
See the difference? The second one has:
– Specific length
– Target audience
– Tone/style guidance
– What to include/exclude
## The Prompt Formula That Works
Every good prompt has these elements:
1. **Role** (optional but helpful): “You are a professional copywriter specializing in SaaS products…”
2. **Task**: “Write a landing page headline for a project management tool…”
3. **Context**: “…targeting freelancers who currently use spreadsheets to track projects…”
4. **Format**: “…in 50 words or less…”
5. **Tone**: “…professional but conversational, no corporate jargon…”
Example:
> “You are an expert financial advisor for working people without finance degrees. Create a checklist of 10 things someone earning $50k-$80k should do before they turn 30. Format as a numbered list. Make it practical (not theoretical). Include one thing that surprises most people.”
That prompt will give you something 10x better than “Give me financial advice.”
## The Advanced Moves
### 1. Examples Help
Instead of telling AI what you want, sometimes showing it works better.
“Here are two examples of good product copy:
[Example 1]
[Example 2]
Now write something in the same style for [your product].”
### 2. Constraints = Creativity
Ironically, more constraints lead to better output.
“Write a Twitter thread (7 tweets max) about why most people fail at investing. Each tweet should be under 280 characters. Use no jargon. No emojis.”
That’s hard to write badly because the constraints are so tight.
### 3. Ask for Reasoning
“Explain your reasoning as you go” often produces better outputs because the AI has to think through its answer.
### 4. Iterate
First prompt rarely nails it. Treat it like a conversation:
– Prompt 1: Get a draft
– Review it
– Prompt 2: “That’s close but make it more [X], less [Y]”
– Repeat
## Common Mistakes
**Too vague:** “Write an article” → AI guesses what you want
**Too long:** 5 paragraphs of instructions → AI loses track
**Contradictory:** “Make it short but comprehensive” → AI can’t win
**No examples:** “Make it sound like our brand” → AI has no reference
## The Real Secret
The best prompt engineers don’t use magic prompts. They:
1. Know what they want (before asking AI)
2. Test different approaches
3. Refine based on results
4. Treat AI like a tool that needs direction
Bad prompt: “Make this better”
Good prompt: “Make this shorter while keeping the main point. Cut it from 500 words to 200.”
## Practical Example
Let’s say you’re writing a product description.
**First attempt:**
“Write a product description for a standing desk”
**Better:**
“Write a 100-word product description for a standing desk that costs $400. Target audience: people who work from home and want to improve their health. Tone: helpful and honest, not salesy. Mention the motor, adjustability, and stability. Include one potential concern someone might have and address it.”
The second will give you something you can actually use.
## What To Practice
1. Pick something you write regularly
2. Spend 2 minutes writing a really specific prompt
3. Compare the result to a generic prompt
4. Notice the difference
You’ll see immediately why prompt engineering matters.
The AI isn’t getting smarter. You’re getting better at telling it what you actually need.
—
Ready to master AI tools? [Start with the basics here](/start-here).