How to Get ChatGPT to Write Better Prompts for You: A Guide Using OpenAI’s Prompt Engineering Cookbook

May 08, 2025

∙ Paid

If you’ve ever used ChatGPT and found its responses a bit off-target or too generic, you’re not alone.

The secret to unlocking ChatGPT’s full potential lies in crafting precise, well-structured prompts.

OpenAI’s Prompt Engineering Guide and Cookbook offer a treasure trove of strategies to help you create prompts that deliver exactly what you need.

I’ll walk you through how to use these resources to write specific, detailed prompts that get high-quality results from ChatGPT.

Let’s dive into the key tactics and examples to make your AI interactions shine.

Why Prompt Engineering Matters

Prompt engineering is the art of designing inputs to guide large language models (LLMs) like ChatGPT to produce relevant, accurate, and tailored responses.

OpenAI emphasizes that “the right prompt can significantly influence the output quality and relevance”.

A vague prompt like “write about marketing” might yield a broad, unfocused response, while a detailed one can generate a laser-focused output that meets your exact needs.

OpenAI’s guide outlines six high-level strategies, and I’ll break down the most actionable ones with practical tips and examples to help you craft better prompts.

Build something Lovable - Lovable is an AI-powered platform that enables users of any skill level to create full-stack web applications without requiring coding expertise by simply describing what they want in plain English.

OpenAI’s Key Strategies for Crafting Specific Prompts

1. Write Clear Instructions

Clarity is the foundation of a great prompt.

OpenAI advises being “specific about your needs” and including relevant details like tone, length, style, or format. Ambiguity leads to misinterpretation, so spell out exactly what you want.

Tactic: Include Specific Details

What to do: Define the task, audience, tone, and output format. For example, instead of “write a blog post,” specify the topic, word count, and style.

Example Prompt:

“Write a 500-word blog post for a tech startup’s newsletter targeting young entrepreneurs. The topic is ‘Why AI Tools Boost Productivity.’ Use a conversational tone, include two real-world examples, and structure it with an intro, three main points, and a conclusion.”

Why it works: This prompt gives ChatGPT a clear goal (blog post), audience (young entrepreneurs), topic, tone, and structure, reducing guesswork.

Tactic: Assign a Persona

What to do: Ask ChatGPT to adopt a specific role or perspective, such as a teacher, marketer, or historian, to tailor the response to a particular expertise or viewpoint.

Example Prompt: “You are a seasoned content marketer with 10 years of experience. Write a 300-word LinkedIn post about the benefits of AI-driven advertising for small businesses. Include one statistic and a call-to-action to spark discussion.”

Why it works: The persona ensures ChatGPT responds with industry-specific knowledge and authority, while the constraints (word count, statistic, CTA) keep it focused.

2. Provide Reference Text

To reduce inaccuracies or “hallucinations,” OpenAI suggests providing reference text or instructing ChatGPT to use citations from a specific source. This grounds the response in reliable information.

Tactic: Use Delimiters for Reference Text

What to do: Enclose reference material in delimiters (e.g., triple quotes ```) to clearly separate it from the instructions.

Example Prompt:

Reference text: “A 2023 study by McKinsey found that AI-driven personalization can increase marketing ROI by 10-20%.”

You are a business consultant. Write a 200-word article section explaining how AI personalization boosts marketing ROI for e-commerce businesses. Cite the McKinsey study and use a professional tone.

Why it works: The reference text provides a factual anchor, and the delimiters ensure ChatGPT knows exactly which information to incorporate.

3. Split Complex Tasks into Simpler Subtasks

Complex prompts can overwhelm ChatGPT, leading to incomplete or scattered responses.

OpenAI recommends breaking tasks into smaller, manageable parts.

Tactic: Use Step-by-Step Instructions

What to do: Outline the task in clear, sequential steps to guide ChatGPT through the process.

Example Prompt:

“I’m creating a social media strategy for a fitness brand. Step 1: Summarize the brand’s goals (increase followers by 20%, boost engagement). Step 2: Suggest three Instagram post ideas with captions. Step 3: Recommend two hashtags for each post. Use a motivational tone.”

Why it works: Dividing the task into steps ensures ChatGPT addresses each component systematically, resulting in a cohesive strategy.

Tactic: Summarize Long Inputs

What to do: If you’re working with lengthy documents, ask ChatGPT to summarize sections before tackling the main task.

Example Prompt:

“Summarize the key points of this 1,000-word report on fitness trends in 100 words. Then, based on the summary, draft a 150-word email to gym owners promoting a new AI workout app. Use a persuasive tone.”

Why it works: Summarizing first helps ChatGPT focus on the most relevant information, improving the quality of the final output.

4. Give the Model Time to “Think”

ChatGPT performs better when prompted to reason through a problem step-by-step, especially for complex or analytical tasks. OpenAI calls this “chain-of-thought prompting”.

Tactic: Request Step-by-Step Reasoning

What to do: Instruct ChatGPT to break down its thought process before providing the final answer.

Example Prompt:

“You are a financial advisor. A client wants to invest $10,000 in a low-risk portfolio. Step 1: List three low-risk investment options. Step 2: Explain the pros and cons of each. Step 3: Recommend one option and justify your choice. Show your reasoning.”

Why it works: Asking for reasoning steps forces ChatGPT to think critically, leading to a more thoughtful and justified response.

5. Use External Tools (When Applicable)

For tasks like calculations or code generation, OpenAI suggests prompting ChatGPT to generate code that can be executed in a sandbox rather than performing the task directly.

This is more relevant for developers but can inspire creative prompt structures.

Tactic: Prompt for Code-Based Solutions

What to do: Ask ChatGPT to write executable code for specific tasks, like data analysis or conversions.

Example Prompt:

“Write a Python function that converts miles to kilometers. The function should prompt the user for a number in miles and return the result in kilometers. Include comments explaining each step.”

Why it works: This leverages ChatGPT’s ability to generate accurate code, ensuring precise results for technical tasks.

6. Test Changes Systematically

Prompt engineering is iterative. OpenAI recommends testing and refining prompts based on the outputs you receive.

If the response isn’t what you wanted, tweak the wording, add more context, or adjust constraints.

Tactic: Iterate and Refine

What to do: Start with a basic prompt, review the output, and revise it to address gaps or inaccuracies.

Example Process:

Initial Prompt: “Write a short story about a haunted house.”

Output Issue: Too generic, lacks setting details.

Refined Prompt:

“Write a 300-word short story about a haunted Victorian mansion in 1920s London. Include a ghostly figure and describe the atmosphere. Use a suspenseful tone.”

Why it works: Iteration hones the prompt to align with your vision, improving specificity and quality.

Practical Tips for Writing Specific, Detailed Prompts

To tie it all together, here are some actionable tips inspired by OpenAI’s guide and cookbook to ensure your prompts are laser-focused:

Be Explicit About Constraints: Specify word count, tone, or exclusions (e.g., “Avoid jargon” or “No bullet points”).

Use Descriptive Adjectives: Words like “formal,” “humorous,” or “concise” guide the tone.

Provide Context: Include background info, such as the project’s purpose or audience, to make responses relevant.

Test Multiple Prompts: Try variations to see which yields the best results, as small tweaks can make a big difference.

Use Examples: Show ChatGPT what you want by including sample outputs or formats.

Real-World Example Prompt:

“You are a travel blogger with expertise in sustainable tourism. Write a 400-word blog post for eco-conscious travelers about the top three eco-friendly destinations in Europe for 2025. Include one specific activity for each destination, use a friendly tone, and structure the post with an intro, three sections (one per destination), and a conclusion. Cite a 2024 report by GreenTravel stating that 70% of travelers prefer sustainable destinations.”

This prompt incorporates a persona, context (2024 report), specific details (three destinations, one activity each), tone, structure, and a reference, ensuring a tailored, high-quality response.

Putting It Into Practice

Now it’s your turn! Start by identifying a task—maybe drafting a newsletter, brainstorming marketing ideas, or writing a story.

Use OpenAI’s strategies to craft a prompt with:

A clear task and objective.
A defined persona or role.
Specific constraints (length, tone, format).
Context or reference text, if applicable.
Step-by-step instructions for complex tasks.

Test your prompt, review the output, and refine it as needed. The more you practice, the better you’ll get at steering ChatGPT to deliver exactly what you envision.

Final Thoughts

Crafting specific, detailed prompts is like giving ChatGPT a roadmap to success.

By leveraging OpenAI’s Prompt Engineering Guide and Cookbook, you can transform vague responses into targeted, high-quality outputs that save you time and effort.

Whether you’re writing for your Substack, creating content for work, or exploring creative projects, these strategies will help you harness ChatGPT’s power like a pro.

Want to share your favorite prompt?

Drop it in the comments or reply to this newsletter—I’d love to see how you’re using these tips!

And if you found this post helpful, subscribe for more AI-powered writing hacks every week.

Download a PDF Document of this guide

Keep reading with a 7-day free trial

Subscribe to ON AI to keep reading this post and get 7 days of free access to the full post archives.