Prompt Engineering Guide: Write Better AI Prompts

📖 12 min read · AI & Machine Learning · Try AI Prompt Builder →

What is Prompt Engineering?

Prompt engineering is the practice of designing and refining inputs to AI language models to get better, more reliable outputs. A well-crafted prompt can be the difference between a vague, generic response and a precise, actionable answer.

Unlike traditional programming where you write exact instructions, prompting is more like communicating with a very capable but literal colleague — the clearer and more structured your request, the better the result.

The Anatomy of a Good Prompt

Every effective prompt has some combination of these components:

Role

Who the AI should act as. Sets expertise and tone.

"You are a senior Python developer with 10 years of experience."

Task

What you want the AI to do. Be specific and use action verbs.

"Write a function that validates email addresses."

Context

Background information the AI needs to give a relevant answer.

"The app uses Django and targets non-technical users."

Format

How you want the output structured.

"Return the result as a JSON object with keys: valid, reason."

Constraints

Limits or rules the AI must follow.

"Keep it under 20 lines. No external libraries."

Examples

Sample inputs/outputs to show the pattern you want (few-shot).

"Input: test@email.com → Output: {valid: true}"

Core Prompting Techniques

1. Zero-Shot Prompting

Ask the model directly without any examples. Works well for simple, well-defined tasks.

Classify the sentiment of this review as Positive, Negative, or Neutral:
"The product arrived on time but the packaging was damaged."

2. Few-Shot Prompting

Provide 2–5 examples before your actual request. Dramatically improves consistency for structured outputs.

Classify sentiment:
Review: "Amazing quality!" → Positive
Review: "Broke after one day." → Negative
Review: "It's okay, nothing special." → Neutral

Review: "Fast shipping but wrong color sent." → ?

3. Chain-of-Thought (CoT)

Ask the model to reason step by step before giving the final answer. Significantly improves accuracy on math, logic, and multi-step problems.

A store sells apples for $0.50 each and oranges for $0.75 each.
If I buy 4 apples and 3 oranges, what's the total cost?
Think step by step before giving the final answer.

4. Role Prompting

Assign a persona to the model. This shifts the vocabulary, depth, and framing of responses.

You are a senior security engineer at a Fortune 500 company.
Review this code for SQL injection vulnerabilities and explain
each issue as if presenting to a junior developer.

5. Structured Output Prompting

Tell the model exactly what format to return. Essential for programmatic use of AI outputs.

Extract the following from this job posting and return as JSON:
- job_title
- company
- required_skills (array)
- salary_range (or null if not mentioned)

Job posting: [paste text here]

System Prompts vs User Prompts

Most AI APIs (OpenAI, Anthropic, Google) support two message types:

Type	Purpose	Example
System	Sets persistent behavior, role, and rules for the entire conversation	"You are a helpful coding assistant. Always include error handling in code examples."
User	The actual request or question for this turn	"Write a Python function to parse CSV files."
Assistant	Previous AI responses (used in few-shot or conversation history)	"Here is the function: def parse_csv(...)..."

💡 Put stable instructions (role, format, constraints) in the system prompt. Put the actual task in the user message. This keeps conversations clean and reduces token usage.

Common Prompt Mistakes

Too vague

✗ Bad: "Write something about Python."

✓ Good: "Write a 300-word beginner-friendly explanation of Python list comprehensions with 3 code examples."

No context

✗ Bad: "Fix my code."

✓ Good: "This Python function throws a KeyError on line 12. Identify the bug and explain why it happens."

Too broad

✗ Bad: "Tell me everything about machine learning."

✓ Good:

"Explain the difference between supervised and unsupervised learning in 2 paragraphs for a developer with no ML background."

Vague constraints

✗ Bad: "Don't make it too long or too short."

✓ Good: "Keep the response between 150–200 words."

Advanced Techniques

→

Self-Consistency: Run the same prompt multiple times and take the majority answer. Useful for factual questions where the model might hallucinate.

→

ReAct (Reason + Act): Ask the model to alternate between reasoning ("Thought:") and actions ("Action:"). Used in AI agents to break down complex tasks.

→

Prompt Chaining: Break a complex task into smaller prompts where the output of one becomes the input of the next. More reliable than one giant prompt.

→

Negative Prompting: Explicitly tell the model what NOT to do. "Do not include disclaimers. Do not use bullet points. Do not repeat the question."

→

Temperature Control: Lower temperature (0.0–0.3) for factual/deterministic tasks. Higher (0.7–1.0) for creative tasks. Not a prompt technique but works alongside it.

Token Awareness

Every word in your prompt costs tokens — and tokens cost money. Key things to know:

•Roughly 1 token ≈ 4 characters or ¾ of a word in English

•Both your input prompt AND the AI's output count toward your token bill

•Longer system prompts are charged on every API call in a conversation

•GPT-4o charges $5 per 1M input tokens — a 1,000-token prompt costs $0.005

•Claude 3.5 Sonnet has a 200K context window — you can send entire codebases

Quick Reference: Prompt Templates

Code Review

You are a senior [language] developer. Review the following code for:
1. Bugs and edge cases
2. Security vulnerabilities  
3. Performance issues
4. Code style and readability

For each issue, explain the problem and suggest a fix.

Code:
[paste code here]

Content Writing

You are an expert content writer specializing in [topic].
Write a [word count]-word [content type] for [target audience].
Tone: [professional/casual/technical]
Include: [specific sections or requirements]
Avoid: [things to exclude]

Data Extraction

Extract the following fields from the text below and return as JSON:
- field1 (type: string)
- field2 (type: number)  
- field3 (type: array of strings)

If a field is not found, use null.

Text: [paste text here]

Build & Test Your Prompts

Use the AI Prompt Builder to structure system prompts, set tone, format, and constraints — then copy directly into ChatGPT, Claude, or your API calls.

AI Prompt Builder →Token Counter →Compare AI Models →