ChatGPT has tremendously impacted AI interactions. Its ability to generate text closely resembling human language makes it useful in content creation and customer support. Using prompts as input, ChatGPT can gather user information and produce novel text samples encompassing natural language, code, and even multimodal elements. Therefore, mastering the art of prompt engineering is crucial for unlocking ChatGPT’s full potential.

To make the most of prompt engineering, it is essential to understand how ChatGPT operates. This model relies on a transformer architecture and leverages extensive text data to produce contextually relevant responses. By crafting well-structured prompts, users can effectively guide the model to generate desired outputs. Prompt engineering represents a revolutionary approach in AI and natural language processing, greatly enhancing language models’ efficiency, accuracy, and safety like GPT-3, LLama, and LaMDA.

This article will explore best practices, tips, and tricks for creating effective prompts while providing real-world examples for illustration purposes.

The Elements of Prompt Engineering

Prompt Engineering

What are the common terminologies used in Prompt Engineering?

  • Tokens: API tokens are byte-sized language units.
  • Zero-shot prompting: Users who directly pose questions without explicitly providing instructions. Although the format and accuracy of responses are limited, this method yields coherent responses.
  • Chain-of-thought prompting: A style of prompting that incorporates a series of intermediate reasoning steps within the prompt. Chain-of-thought prompting encourages the model to reason in a step-by-step fashion.

Be Clear and Precise

When it comes to formulating prompts for ChatGPT, clarity and precision are of utmost importance. It is crucial to ensure that your prompt is concise and effectively conveys the information you seek from ChatGPT. Avoid using vague language and provide sufficient context for the model to understand your request.

For example, instead of using a vague prompt like “Tell me about AI,” it is better to use a specific and clear prompt such as “Explain the key principles of artificial intelligence and its applications in healthcare.”

Let’s consider a practical scenario where a medical researcher wants insights into recent developments in AI in diagnostic medicine. A valuable and specific prompt could be: “Provide a summary of the latest breakthroughs in AI-driven diagnostic tools designed for cancer detection.”

Leverage System Messages to Enhance Context

Providing adequate context is crucial for generating high-quality responses in the model. It is essential to supply sufficient context to ensure that the prompts are relevant and accurate. One practical approach when using ChatGPT is incorporating system messages into the conversation. You can effectively guide and direct the model’s behaviour throughout the interaction by including a concise system message at the beginning. This method enhances the production of precise and meaningful responses.

For Example:

    System message: “You are an assistant that provides information about renewable energy sources.”

    In practical scenarios, a marketing team working on a campaign for an electric vehicle company could use a system message such as:

    System message: “You are an assistant that offers insights on the benefits of electric vehicles and their impact on the environment.”

    This approach ensures that ChatGPT aligns with the intended context and objectives of the conversation.

    Get Creative with Prompt Styles

    Embrace the art of experimentation regarding prompt formats, as they can make a difference. Try various styles like questions, statements, or instructions to find one that perfectly suits your needs.

    You might find these examples inspiring:

    • Question: “What are the benefits of solar energy?”
    • Statement: “Discuss the advantages of solar energy.”
    • Instruction: “List the top five benefits of solar energy.”

    In real-world scenarios, imagine you’re a content writer seeking ideas on topics for sustainable fashion:

    • Question: “How does sustainable fashion contribute to environmental conservation?”
    • Statement: “Describe the role of sustainable fashion in reducing the fashion industry’s environmental impact.”
    • Instruction: “Provide three examples of sustainable fashion brands and their eco-friendly practices.”

    By exploring different prompt styles, you can uncover the one that perfectly aligns with your objectives and sparks the engagement you’re looking for.

    Tailor Response Length

    To manage ChatGPT’s response length, you have two options: set a character limit or explicitly specify the desired size in your prompt.

    Example scenarios:

    • Character Limit: You can limit responses to 280 characters for a concise tweet.
    • Explicit Request: You may request a summary of an article in just three sentences.

    To concisely announce a new product, a social media manager may set a character limit of 280 and pose the question: “How would you describe the key features of our latest smartphone in a tweet?”

    Determine the  Output Format

    In addition to controlling the response length, you can also determine the format in which ChatGPT provides its output. By specifying whether you prefer JSON or XML, for example, you can have greater control over the structure of the response.

    Sample prompts with specified output format:

    • “Suggest five top veg restaurants near the Eiffel Tower in Paris, and format the response as a JSON string.”
    • “Recommend fiveself-helpp books in XML format.”

    From a practical standpoint, a book recommendation application could utilise this methodology to seamlessly incorporate curated lists of books into its backend infrastructure, thereby optimising the overall development process of the app.

    Go for Positive Instructions

    When crafting prompts, focus on positive instructions. Instruct the model on what to do instead of specifying what not to do. This approach minimises confusion and enhances the model’s ability to generate desired outputs.


    • Less Effective: “Exclude negative reviews from the product description.”
    • More Effective: “Include only positive reviews in the product description.”

    In a practical scenario, an e-commerce company may utilize a constructive prompt such as: “Craft a product description that emphasises the product’s strengths and positive customer feedback,” guaranteeing that the generated responses effectively promote the product’s merits rather than its shortcomings.

    Another instance:

    • Less Effective: “Avoid using technical jargon in the response.”
    • More Effective: “Use plain language to explain the concept.”

    To help customers gain clarity, a financial services firm could utilise a positive prompt such as: “Please elucidate the concept of compound interest in simple language.” 

    Incorporate Contextual Insights

    Enhance prompts by incorporating contextual information to elicit more precise responses. To generate customised results, provide user-specific details such as location, preferences, or past interactions.


    • Contextual Prompt: “Recommend Italian restaurants near the New York Empire State Building that offer vegetarian options.”

    Imagine using a travel app that considers your exact location and dietary preferences to give you personalised restaurant recommendations. This innovative technology aims to provide a tailored experience, ensuring users can find exactly what they want, even in unfamiliar places.

    Iterate and Refine

    Prompt engineering thrives on iteration. Experiment with various prompts, evaluate outcomes and refine your approach based on the model’s performance. Persistence is critical to achieving your desired results.

    In the real world, a customer support team using ChatGPT for FAQs might test diverse prompts to optimise responses. For example:

    • Initial Prompt: “How do I reset my password?”
    • Refined Prompt: “Provide step-by-step instructions for resetting a forgotten password on our platform.”

    Through continual refinement, you can enhance the effectiveness of your prompts and meet evolving needs efficiently.


    Mastering prompt engineering is a multi-faceted journey combining expertise in natural language processing, machine learning, and programming, focusing on languages like Python and frameworks like TensorFlow. Proficiency in data structures, algorithms, data cleaning, and preprocessing is fundamental. Furthermore, a clear grasp of the tasks for which Large Language Models (LLMs) are trained, such as sentiment analysis or text summarisation, is essential. Effective communication skills are vital, given that prompt engineering often entails collaborative efforts.

    In addition, a deep understanding of the domain you operate in and the preferences of your target audience is crucial for crafting prompts that truly resonate. Prompt engineering serves as a potent tool to amplify the capabilities of LLMs, offering precise input-output pairs to guide them in producing accurate and efficient outcomes. Techniques like zero-shot prompting, few-shot prompting, chain-of-thought prompting, self-consistency generate knowledge prompting, active-prompt directional stimulus prompting, and multimodal graph prompting empower LLMs to excel with limited training data, generate innovative outputs, and adapt to novel tasks seamlessly.