Master the Art of Prompt Writing

Ever found yourself staring at ChatGPT, typing “help me with…” and getting responses that are… well, not quite what you had in mind? You’re definitely not alone. Here’s the thing: most of us treat AI like it’s a mind reader. We throw vague questions at it and expect magic. But here’s what I’ve learned after countless hours of prompt experimentation, the quality of your question determines the quality of your answer.

In this article, I’m going to walk you through everything I wish someone had told me when I started working with AI tools.

# 💡 What you’ll discover:

• Why your current prompts probably aren’t working (and it’s not your fault)
• The psychology behind how LLMs actually “think” and make decisions
• A simple 4-step process that transforms vague ideas into crystal-clear prompts
• Real before/after examples that’ll make you go “ohhhh, THAT’S why it wasn’t working”
• A practical checklist you can use immediately

Fair warning: This isn’t about quick hacks or magic formulas. It’s about understanding the fundamentals so you can write prompts that actually get you what you need, every single time._ Ready to stop fighting with AI and start collaborating with it? Let’s dive in.

# **The key to writing good prompts lies in:CLEAR and EASY texts

To achieve that, you must clear your mind first, so write in a draft file:

• What you know
• What you need to know
• What do you think is the way of knowing it

The main error is thinking that the LLM already knows what you want,
But first, you need to know what you want.

# LLMs tend to pick shortcuts,

This is very important to understand, LLMs use tokens as a basic unit.
They tend to be cost-efficient, so they will try to answer your question with the smallest token possible. Unless you prompt them to “think” and write down the whole argument they can come up with.

We can say that LLMs are cost-efficient, as our bodies are.
Usually, we don’t think about the “right”:

• Amount of water to drink
• The number of hours to work out

We just drink and work out as we feel it, unless you set your goals for working out.
In other words, you don’t think about working out unless you are explicit about what you want!

Think about this concept, also re-read this paragraph if needed.

# Applying the cost-efficient concept to prompts:

When you ask a general question, there are endless ways of answering it, as well as endless possible shortcuts. The more specific and clear you are, the better the answer will be.

# To be specific, you must:

1. Include any relevant info about
2. Organize them
3. Clearly state your intentions
   1. What is the info about?
   2. How does the AI need to use them?
   3. What’s the goal?

# 💡NOTE:

English is full of polysemic words, so one word may mean multiple things based on the context. After you write any prompt ask yourself:

• are there any ambiguities?
• if so, how can I remove them?

Most of the time, you will need to rewrite your prompts again and again, cause the high-level process is:

1. Think
2. Write
3. Try it to get feedback, go back to step 1

Of course, this is a way to get relevant and right information, so you should understand that:

• In the short term
  • You will need more time
  • to find the right question
• But in the medium/long period, * You will have a deep understanding * of any topic you were curious about

This was the premise for entering the prompt writing state of mind. Let’s now explore how to organize the prompt!

# Text Structure

Text Structure is critical.
You must structure the text easily. When you read it, it must flow by itself.
The key point to master is: SHAPE IS CONTENT For example, titles are the pillars of a clear text. A good title describes the whole paragraph.
It lets the reader know what he’s going to read to tune his mind accurately.

The best way to structure a text is in Markdown format. More on this here
The base layer of the text must be in Markdown, use lists, and hierarchical titles.
Let’s explore both:

# Lists

Note that lists are useful for LLMs (as well as for humans) because they increase readability and make you think more clearly. There are 2 types of lists:

# Numbered list

Numbered lists are perfect when the LLM needs to follow a step-by-step process. You can easily describe the process logically and sequentially. Markdown example:

1. Read the following data: {data}
2. Extract and write the most relevant infos about {topic}
3. Answer the user's question: {question}

# Bullet list

Bullet lists are useful for describing a group of entities. When you cite multiple elements in the same context, use bullet lists. Markdown example:

This data contains 3 main infos:
- identification code
- description
- language used

# Hierarchical Titles

A good title describes the whole paragraph.
It lets the reader know what you are going to read to tune their mind accurately. Hierarchical Titles describe the whole text structure, allowing you to define the importance and specificity of paragraphs.

• The more you dive deep into something, the thinner the title.
• Big titles inform about the main topics.
• Smaller titles provide specific info.

Markdown Example:

# Vehicle
Vehicles are a general category that includes ...

## Motorcycle
Motorcycles are a specific type of Vehicle that has 2 wheels and ...

#### Italian Motorcycles
In Italy there are many Motorcycle brands, the best one is ...

# How to write Prompts?

So now we discussed the tool needed to write a prompt, but we didn’t disclose yet HOW to write it.

Till now…

The main concept to think about is finding the:

• right question or
• right perspective

As we said at the start of the document, the LLM doesn’t know yet what you need. The first thing is discovering it by yourself or using the LLM. We are entering the topic of how to approach a problem.
The best way consist of 3 steps:

# 1. Gather Process

Gather info about the context

1. Understand in which context the question and answer live.
2. For example, are we talking about Computer Science or Literature?
3. For example, are we talking about how to write prompts or how to think and elaborate thoughts?

# 2. Guess Process

Form your intuition

1. After you’ve got some info, you can guess a potential output
2. Honestly, it doesn’t matter how close you get to the correct answer
3. The important thing is to TRY
4. Annotate your guess in a separate file and forget about it for now
5. This step is crucial because it is the point where you begin to improve
6. You will get feedback based on this guess

# 3. Formulation Process

Formulate the question

1. Based on the info learned from Gather Process and Guess Process
2. You now have some experience in the given context
3. So you are now ready to write a proper question.

The steps should finish here, BUT they don’t.

# 🤝 Let’s connect on LinkedIn and tell me what you think!

If you noted, the first step, Gather Process , consists of asking questions…
It might seem a little tricky, but the actual flow is the following:

Until I got the right Question:

1. Gather Proces
2. Guess Process
3. Formulation Process
4. Check Process: 1. Ask yourself:
   1. Did the answer resolve my problem?
   2. Does the result satisfy me?
   3. Do I reached the goal?

 2. If yes: **congratulations, you made it**
 3. If no:
    1. `Compare your guess of the step 2 with the answer you got`
    2. `Analyze the info you now have`
    3. `Start again from step 1 Gather Process, with much more infos!`

Of course, this is not simple at all for at least 3 reasons:

1. You don’t know if you made it until you do, and it might get a little stressful.
2. This is not a linear process:
   1. We are prone to think linearly, but actually a lot of processes are not linear at all.
3. We live in the Dopamine Nation:
   1. Most people would say: “I use ChatGPT cause I don’t want to do full research.” It is easy to think there’s an entity that knows everything and everything is ready right away, but that’s not the case even for LLMs.

#

Prompt Writing Checklist Next, we will see 3 examples with real use-cases, but first, save this checklist to remind yourself how to write good prompts!
Before submitting your prompt, check:

• Context clearly stated
• Specific goal identified
• Ambiguous terms eliminated
• Format requirements specified
• Expected output defined

# 💡Key-Take-Aways:

1. The most important thought is, "Did I formulate the `Right` question? "
2. Only when you own the context you can formulate the `Right` question.
3. It takes time and effort.

# Examples

Let’s analyze three real-world scenarios showing the evolution from basic to effective prompts. The following provides a Prompt and a Result, but I encourage you to copy and try both prompts!

# Example 1: Writing a Blog Post

❌ Ineffective Prompt:

Write a blog post about AI

✅ Effective Prompt:

Write a blog post about the impact of AI on small businesses in 2024.
Target audience: Small business owners
Key points to cover:
- Cost-effective AI tools for automation
- Implementation challenges and solutions
- ROI examples from real businesses
Tone: Professional but approachable
Length: 1000-1200 words
Format: Include H2 headings, bullet points, and a conclusion with actionable steps

Why it works: Notice how the improved version provides clear context, specific requirements, and most importantly, state a clear goal.

# Example 2: Code Review

❌ Ineffective Prompt:

Check this code for errors:
def authenticate_user(username, password):
    db = mysql.connector.connect(
        host="localhost",
        user="root",
        password="123456",
        database="users"
    )
    cursor = db.cursor()
    query = "SELECT * FROM users WHERE username='%s' AND password='%s'" % (username, password)
    cursor.execute(query)
    user = cursor.fetchone()
    if user:
        session['user'] = username
        return True
    return False

✅ Effective Prompt:

def authenticate_user(username, password):
    db = mysql.connector.connect(
        host="localhost",
        user="root",
        password="123456",
        database="users"
    )
    cursor = db.cursor()
    query = "SELECT * FROM users WHERE username='%s' AND password='%s'" % (username, password)
    cursor.execute(query)
    user = cursor.fetchone()
    if user:
        session['user'] = username
        return True
    return False

Review this Python authentication function for:
1. Performance optimization opportunities
2. Security vulnerabilities
3. Code style compliance with PEP 8
Context:
- This function processes user authentication in a web application
- It needs to handle sensitive user data
- It will be used in a production environment
- We need to follow OWASP security guidelines
Expected output:
- List specific security issues found
- Provide code snippets showing recommended fixes
- Explain the reasoning behind each suggestion
- Include best practices for secure authentication
Format:
- Organize findings by severity (Critical, High, Medium, Low)
- Include example of secure implementation
- Add comments explaining security measures

Why it works: The prompt provides specific review criteria, security context, and clear expectations for the output format. It guides the AI to focus on critical security aspects of authentication.

# 💡Learn more about AI and Automations by sending a DM on LinkedIn !

# Example 3: Market Research

For this example, we will use Perplexity, as it can retrieve updated information. ❌ Ineffective Prompt:

Tell me about the market for mobile apps

✅ Effective Prompt:

Analyze the mobile fitness app market with focus on:
Target market: Working professionals (30-45 years old)
Geographic region: North America
Time period: Last 2 years
Please provide:
1. Top 3 current market trends with supporting data
2. Key user pain points from app reviews
3. Revenue models of successful competitors
4. Potential market gaps for new entrants
Format the analysis in bullet points with relevant statistics.

Why it works: The prompt defines clear parameters and specific deliverables.

# Advanced Prompting Preview

More sophisticated techniques include:

• Use AI to improve your prompts
• Chain-of-thought prompting
• Role-based context setting

Want to master these advanced techniques? Book a free consulting call to learn how to:

• Improve response accuracy
• Implement advanced prompt chaining
• Create custom tools for your specific needs

# Conclusion

As you noted, these prompts are very linear, structured, and simple. They describe the situation clearly.

Being Simple is the key to everything.

Let’s connect on LinkedIn, send me a DM if you’re interested in a free consulting call!

# Thanks for reading! 👋