TOPIC / PROMPT ENGINEERING

WHY YOUR PROMPTS SUCK

Most people type random thoughts into ChatGPT and wonder why the output is mid. The difference between a $0 response and a $200/hr consultant response? How you ask.

{?.?}
Bad Prompt
Write me a marketing email.
Good Prompt
You are a conversion-focused email copywriter. Write a launch email for a new AI productivity app targeting freelancers aged 25-40. Format: Subject line + 150-word body. Tone: Casual, direct, no corporate speak. Include: One clear CTA, a pain point hook, social proof placeholder. Avoid: Buzzwords like "synergy" or "leverage."
01 / THE PROBLEM

WHY MOST PEOPLE WRITE TERRIBLE PROMPTS

Here's the thing nobody tells you: AI models are people-pleasers. If you give them a vague prompt, they'll give you a vague answer and act like it's exactly what you wanted. They won't push back. They won't ask clarifying questions (unless you tell them to). They'll just... generate something mid and call it a day.

90%

of prompts are one sentence with zero context

3x

better output when you assign a specific role

70%

of users never iterate on their first prompt

The problem isn't the AI. The problem is you're talking to it like a search engine instead of treating it like a brilliant intern who needs clear instructions. You wouldn't hand a new hire a task that just says "do marketing." You'd give them context, expectations, and examples. Same rules apply here.

02 / THE FRAMEWORK

THE ANATOMY OF A GOOD PROMPT

Meet CRAFT -- the five elements that turn garbage prompts into gold.

C
Context
R
Role
A
Action
F
Format
T
Tone / Constraints
C — Context

Give the AI background info. What's the situation? Who's the audience? What do they already know?

EXAMPLE
I'm a freelance web developer pitching to small business owners who don't know tech jargon.
03 / COMMON MISTAKES

5 PROMPT MISTAKES YOU'RE DEFINITELY MAKING

Before and after examples. No sugarcoating.

BEFORE
Write me something about marketing.
AFTER
You are a B2B content strategist. Write a 500-word LinkedIn post about why cold email outreach is dying in 2026. Target audience: startup founders. Tone: data-driven but conversational. Include 2 specific stats.
THE FIX

Add who, what, how long, for whom, and what style.

04 / LEVEL UP

ADVANCED TECHNIQUES

From basic to boss-level. Each technique unlocks a new tier of output quality.

Zero-Shot
Basic

Just ask directly. No examples. Works for simple, well-defined tasks where the AI already knows what you want.

Translate 'good morning' to Japanese.
Few-Shot
Intermediate

Give 2-3 examples of what you want, then ask for more. The AI learns the pattern from your examples.

Classify these customer messages: Message: "My order never arrived" -> Category: Shipping Message: "The button doesn't work" -> Category: Bug Message: "Can I get a refund?" -> Category: Billing Now classify: "How do I upgrade my plan?"
Chain-of-Thought
Intermediate

Force the AI to show its reasoning step by step. Add "think step by step" or "explain your reasoning." Dramatically improves accuracy on complex problems.

A store has 15 apples. They sell 7 in the morning and receive a shipment of 12 in the afternoon. Then they sell 5 more. How many apples do they have? Think step by step before giving the final answer.
System Prompts
Advanced

Set a persistent identity and behavior that lasts the entire conversation. This is how you build custom AI assistants with consistent personalities.

[System] You are CodeReview-Bot, a senior staff engineer. You review code with: - Focus on bugs, not style preferences - Security vulnerabilities flagged as [CRITICAL] - Performance suggestions labeled [PERF] - Always explain WHY something is a problem - Be direct but not condescending
05 / TEMPLATES

STEAL THESE PROMPT TEMPLATES

Copy, paste, fill in the brackets. Instant upgrade.

Writing
Role: You are a [type] writer with expertise in [topic]. Task: Write a [format] about [subject]. Audience: [who is reading this] Tone: [casual/formal/provocative/etc.] Length: [word count] Constraints: - [what to include] - [what to avoid] - [style notes]
Coding
Role: Senior [language] developer. Task: [Write/Debug/Refactor/Review] code that [specific behavior]. Context: This is for a [type of project] using [framework/stack]. Requirements: - [Specific behavior 1] - [Specific behavior 2] - Include error handling - Add comments explaining non-obvious logic Format: Return only the code in a code block. No explanations unless I ask.
Research
Role: Expert research analyst in [field]. Task: Research [topic] and provide a structured summary. Format: 1. Key findings (3-5 bullet points) 2. Data/evidence supporting each point 3. Opposing viewpoints or limitations 4. Sources or recommended reading Constraints: - Cite specific data points where possible - Flag anything that might be outdated (pre-2025) - Distinguish between established facts and emerging trends
06 / PRACTICE

YOUR TURN — FIX THESE PROMPTS

Open ChatGPT or Claude right now. Take each bad prompt below and rewrite it using what you learned.

1
Improve this prompt
BAD: Write a blog post about AI.

Hint: Add a role, specific topic, audience, format, length, and tone constraints using the CRAFT framework.

2
Add chain-of-thought
BAD: Is this a good business idea: an app that matches dog owners for play dates?

Hint: Add 'Analyze this step by step considering: market size, competition, revenue model, user acquisition, and potential risks.'

3
Use few-shot prompting
BAD: Rewrite this email to sound more professional.

Hint: Provide 2-3 before/after examples of casual-to-professional rewrites, THEN give it the email you want rewritten.

07 / FAQ

FREQUENTLY ASKED QUESTIONS