Prompt engineering is a crucial skill for making the most out of large language models like GPT-4. It involves crafting input queries in a way that optimizes the model’s understanding and increases the likelihood of receiving useful and coherent responses. This instructional guide will teach you the basic principles of prompt engineering, along with tips and techniques to help you effectively communicate with GPT-4.

Who knows more about prompt engeneering than GPT-4 itself?

This prompt engineering guide teaches you the principles and techniques needed to craft effective prompts, including specificity, leading with questions, and more.

  1. Be specific: GPT-4 is more likely to provide relevant and accurate answers if your prompt is clear and specific. Avoid ambiguity and provide necessary context to guide the model towards a useful response.
  2. Lead with questions: Starting your prompt with a question encourages GPT-4 to provide a direct answer. This can be especially helpful when seeking factual information or advice.
  3. Make it conversational: GPT-4 is trained on a wide range of text, including dialogues. By framing your prompt as a conversation, you can encourage more natural and contextually appropriate responses.
  4. Limit the scope: Break down complex questions into simpler parts. GPT-4 may struggle to provide a comprehensive answer to a broad question, so consider splitting it into smaller queries to receive more accurate information.

Prompt Engineering Techniques

Systematic prompting:

Use a step-by-step approach to refine your prompt, beginning with a simple query and gradually adding more context or specificity. This technique can help identify the best way to communicate your question or request to GPT-4.


  • Start: “Tell me about AI.”
  • Refine: “What are the primary applications of artificial intelligence?”
  • Final: “Can you explain the role of artificial intelligence in healthcare, specifically in diagnostics and treatment?”

Prompt chaining:

Chain multiple prompts together to create a more in-depth conversation. This allows GPT-4 to maintain context and build upon previous responses.


  • Prompt 1: “What is the greenhouse effect?”
  • Prompt 2: “How do human activities contribute to the greenhouse effect?”
  • Prompt 3: “What are some strategies to mitigate the negative impacts of the greenhouse effect?”

Conditional instructions:

Use conditional statements to guide GPT-4’s response. This technique can help you obtain a more focused answer by specifying certain conditions or constraints.


“Explain the process of photosynthesis, but do so in a way that a 10-year-old could understand.”

Answer format:

Specify the desired format of the answer, such as a list or a summary. This can help GPT-4 structure its response in a way that best suits your needs.


“List five renewable energy sources and provide a brief description of each.”

krautart The Art of Prompt Engineering b185f410 e63b 472b 9f9c ea55a1cda3db e1681708441315


If GPT-4’s response is not satisfactory, consider the following steps to refine your prompt:

  1. Reevaluate context: Ensure that you have provided enough context or background information for GPT-4 to understand your question. If the model seems confused, try rephrasing or adding more context.
  2. Adjust specificity: If the response is too broad or not focused enough, make your prompt more specific. Conversely, if the response is too narrow, consider broadening the scope of your question.
  3. Experiment with phrasing: Try rephrasing your prompt or using synonyms to see if it improves the response. The model may better understand certain phrasings based on its training data.
  4. Iterate: Don’t be afraid to experiment and iterate on your prompts. As you learn more about GPT-4’s strengths and weaknesses, you can refine your approach to prompt engineering.
krautart The Art of Prompt Engineering 7dd56eef ddeb 4601 979a b5c48f476e48


Prompt engineering is a vital skill for optimizing your interactions with GPT-4. By applying the principles and techniques outlined in this guide, you can enhance the model’s ability to understand and respond to your queries effectively. Remember to be specific, lead with questions, make your prompts conversational, and limit the scope. Leverage systematic prompting, prompt chaining, conditional instructions, and answer formats to achieve better results. Finally, troubleshoot and iterate by reevaluating context, adjusting specificity, experimenting with phrasing, and learning from your interactions with GPT-4. With practice, you will become adept at crafting prompts that elicit high-quality responses and unlock the full potential of GPT-4.

This text was written by chatGPT und is the first part of a prompt engineering course from beginner to master.

Watch out for the next parts of our series!