Master ChatGPT Prompts in 6 Easy Steps (Step-by-Step Guide)

Struggling with AI that doesn't quite 'get' what you're after? It's likely not the AI's fault – it's the prompts. Let's fix that.

This approach is not my own, I’m found it via The Rundown AI. I’ve been using it ever since. Check out their newsletter for all things AI.

The Problem: Vague Prompts Lead to Vague Results 
When prompts are too broad, they often result in answers that don't quite fit what we need—like a key that doesn't match the lock. For marketers, this means extra hours spent sifting through responses that miss the mark.

Let's simplify the process and get straight to the answers that work for you.

The Solution: Six Ingredients for Perfect Prompts:

1. Task

2. Context

3. Examples

4. Persona

5. Format

6. Tone

Here's a sneak peek at our secret sauce, but keep reading for the full breakdown:

1. Task

This is where you tell ChatGPT what to do. Start with a clear action verb and outline your goal. Be direct and specific.

Why it's important: It sets the direction and ensures ChatGPT knows exactly what you're asking for.

How to use it:

• Simple: "Generate a list of low-carb breakfast options."

• Complex: "Create a detailed marketing plan for a boutique hotel targeting eco-conscious travelers."

2. Context

Context is the setting for your task. It's the who, what, when, where, and why that frames your request.

Why it's important: It gives ChatGPT the necessary background to tailor its response to your specific needs.

How to use it:

• "I'm a vegan runner needing high-protein meal ideas that are quick to prepare."

• "Our hotel is in a tropical location and appeals to travelers who prioritize sustainability."

3. Examples

Examples are your way of showing ChatGPT the format, style, or approach you want it to use.

Why it's important: It gives the AI the structure and tone you're looking for, moving away from generic responses.

How to use it:

• "Write a product description like the one on Apple's website for our new Apple Watch."

• "Draft an apology email similar to the one JetBlue sent to customers after the 2017 flight cancellations."

4. Persona

Persona is about assigning a character or role for ChatGPT to play. It could be an expert, a celebrity, or a fictional character.

Why it's important: It adds depth and perspective to the response, making it more authentic and tailored.

How to use it:

• "You're a seasoned travel blogger. Write a post about hidden gems in Lisbon."

• "Pretend you're Gordon Ramsay reviewing a high-end restaurant's new menu."

5. Format

The format dictates how you want the information presented. It could be a list, a table, a detailed report, etc.

Why it's important: It organizes the output in a way that's most useful to you.

How to use it:

• "List the pros and cons of remote work in bullet points."

• "Summarize the quarterly sales data in a table format."

6. Tone

Tone sets the mood of the output. It could be professional, friendly, sarcastic, etc.

Why it's important: It ensures the response not only fits the content but also the emotional context and audience.

How to use it:

• "Write a witty response to a customer's negative review of our app."

• "Draft a professional yet empathetic response to an employee requesting leave."

By following these steps, you'll transform your prompts from vague to precise, ensuring ChatGPT delivers exactly what you need.

BONUS: Master Prompt Template: 

If you don’t want to start each prompt from scratch, use this master template:

Forget all the previous instructions. Act as a professional {PERSONA}.

You will do the following:
{TASK, STEP BY STEP IF NEEDED}

These principles and rules are important: {CONTEXT}.
These are example(s) of output I expect: {EXAMPLES}
Answer in the following format: {FORMAT}
Use this Tone of Voice: {TONE OF VOICE}

Take a deep breath and work on this problem step-by-step."

Thanks for reading!

PS: last email I asked you to help me improve this newsletter, but I totally forgot to make the form public. So for a second round, please help me improve this newsletter: