Basic Prompting: How to Ask Good Questions
The Single Biggest Mistake Beginners Make
They treat Claude like a search engine. They type short, keyword-style queries and get mediocre results, then conclude "AI isn't that useful."
The truth: the quality of Claude's output is directly proportional to the quality of your input.
This lesson will give you a repeatable framework for writing prompts that get excellent results.
The Anatomy of a Good Prompt
A strong prompt usually contains some combination of these elements:
| Element | What It Is | Example |
|---|---|---|
| Role | Who Claude should be | "You are a senior product manager..." |
| Task | What you want | "...write a one-page brief..." |
| Context | Relevant background | "...for a new feature we're adding to our invoicing app. Our users are small business owners." |
| Format | How you want the output | "Use bullet points. Keep it under 300 words." |
| Constraints | What to avoid | "Don't use jargon. Assume no technical background." |
You don't always need all five. But the more of these you include, the better the result.
Before and After Examples
Example 1: Email Writing
Before (vague):
"Write an email about the project delay"
After (specific):
"I need to email my client Sarah to let her know that our website redesign project will be delayed by 2 weeks. The reason is that we're waiting on content from her team. Keep the tone professional but warm. Don't make it sound like it's her fault, even though it is. End with a clear ask for a new timeline."
The second prompt will produce something you can actually send.
Example 2: Explaining Something Complex
Before:
"Explain databases"
After:
"Explain what a database is to someone who has never coded before. Use an analogy to something from everyday life. Keep it under 150 words."
Example 3: Brainstorming
Before:
"Give me business ideas"
After:
"I'm a nutritionist with 8 years of client experience. I want to create a digital product that generates passive income. My audience is busy professionals aged 30–50 who want to eat better but don't have time to cook elaborate meals. Give me 5 specific product ideas with a one-sentence description of each."
The "One Shot vs. Iteration" Mindset
Sometimes you'll get exactly what you need on the first try. Often, you won't — and that's fine.
Think of prompting as a conversation, not a command. Your first message sets direction; follow-up messages refine it.
A typical iteration looks like:
- First prompt: describe what you want broadly
- Claude responds
- You say: "Good, but [specific change]"
- Claude revises
- Repeat until you're happy
This is faster than trying to craft the perfect prompt on the first attempt.
Useful Prompting Patterns
The "Give me options" pattern
"Give me 3 different versions of this headline, ranging from conservative to bold."
The "Explain like I'm 10" pattern
"Explain [complex concept] as if you're talking to a curious 10-year-old."
The "Act as an expert" pattern
"Act as an experienced UX designer. Review this landing page copy and tell me what's working and what should change."
The "Devil's advocate" pattern
"I'm planning to [decision]. What are the strongest arguments against doing this?"
The "Step by step" pattern
"Walk me through how to [task] step by step. Assume I have no prior experience."
The "Summarize" pattern
"Here's a long document. [paste document]. Give me: 1) A 3-sentence summary, 2) The 5 most important points, 3) Any action items or decisions required."
When Claude Misunderstands You
If Claude's response is way off from what you wanted, it usually means your prompt lacked context. Don't just repeat the same prompt. Instead:
- Tell Claude what was wrong: "That's not what I was looking for."
- Add more context: "I should have mentioned that..."
- Ask Claude to try again: "Please try again with that in mind."
You can also ask Claude directly: "What information would help you give me a better answer?"
A Word on Honesty and Accuracy
Claude is trained to be honest, but it's not infallible. Some rules of thumb:
- Trust Claude's reasoning, but verify specific facts (dates, statistics, names)
- Trust Claude's writing quality — it's almost always good
- Be skeptical of very specific numbers Claude produces without a source
- When in doubt, ask Claude to explain its reasoning: "How confident are you in that, and how would I verify it?"
Summary
- Good prompts include: role, task, context, format, and constraints
- Be specific — vague prompts get vague answers
- Iterate rather than trying to write the perfect prompt first
- Use patterns: options, ELI10, expert role, devil's advocate, step-by-step
- Verify specific facts; trust Claude's writing and reasoning