Does the way you write a prompt really make a difference when you’re interacting with an AI tool, like ChatGPT? The short answer is, yes.
A lot of factors contribute to the response you get from an AI tool or chatbot, including the way it was initially created and the data it’s been “trained” on. But one of the most important factors is how you communicate with it – this is where prompting comes in.
A prompt is essentially a question or an instruction that you give an AI tool to generate a response. It can be a short and simple query, a longer, more structured request, or even a series of guidelines, all with the aim of shaping the output and, ultimately, getting what you want. Which means the way you write these prompts directly impacts how clear, relevant, and useful the AI’s response is.
When I first tried using ChatGPT, I spoke to it like a friend, firing off casual and frantic questions and expecting intuitive responses. While there’s nothing necessarily wrong with this approach, I soon realized that being more intentional about how I wrote my prompts gave me much better answers. This is what’s called ‘prompt engineering’, a growing field that’s become so important, people are building careers out of it.
The science of prompts
Everyone has a slightly different approach when it comes to what makes a great prompt – and this can vary depending on what you need and what tool you’re using.
But, generally, the effectiveness of AI responses has three core elements to consider. The first is context, so an AI tool will perform at its best when it knows enough background to understand what you’re asking it. The more relevant information included, the more accurate and useful the response is likely to be.
Then there’s specificity, so the more clear and precise your prompt is, the better your result will be. Vague prompts tend to produce generic or less relevant answers
Finally, consider structure. If you put time into thinking how you’re wording the prompt, what order you’re shaping it, whether you’re using bullet points or numbers to divide up text, etc, you’re likely to get a more relevant response.
It’s also worth understanding how the AI tool you’re using actually works. For example, chatbots that use Large Language Models (LLMs), which includes ChatGPT, Claude and Google Gemini, generate their responses largely based on probability.
Which means they predict the likely sequence of words based on the input you’ve provided. So any changes you make to the way you write a prompt, even if you think it’s generally asking the same sort of thing, could deliver really different results.
Of course, some people have exaggerated the complexity of prompt engineering to sell courses or consultancy services. However, research shows that a well-constructed prompt does yield better AI-generated results. For example, one study found that a well considered prompt can increase the quality of an LLM response by 57.7% and the accuracy by 67.3%.
But you don’t need a research paper to see this for yourself. Try using your preferred AI chatbot to answer a question using a short, vague prompt –– then refine it with more details and instructions. I bet you’ll be surprised by how much the response improves.
How do you write better AI prompts?
People have conducted research, created courses and made lots of video content to help you demystify the perfect prompt.
But that doesn’t mean there aren’t some key lessons beyond context, specificity and structure to consider. Here’s some of the most popular advice:
- Give examples: In prompt engineering, this is called few-shot prompting. It involves providing the AI with examples before asking it to generate a response. For instance, if you want a well-structured list of travel tips, you might first present two or three sample tips and then ask the AI to generate more in the same style.
- Specify the audience: Define who the content is for. For example, if you’re asking for skincare advice, specify whether it’s for dermatologists, beauty enthusiasts, or time-poor individuals with acne. This helps the AI tailor its response.
- Explain who you are: As well as specifying who the output is for, tell the AI tool who you are can make the output more relevant. For example, saying “I’m a teacher creating a quiz for my students” will lead to a more structured, educational response than a generic request for quiz questions.
- Tell it the format you want: Make it clear whether you want the output to be presented as an article, script, bullet point list or guide. You can also specify the word count limit to fine-tune the response.
- Define the tone: If you’re writing for a specific purpose, give details about the tone you want. Defining the style upfront helps refine the AI’s output. You can also provide examples or links to reference materials for better accuracy.
- Use natural language: While structured prompts are important, over-complicating them can backfire. AI models are designed to interpret natural language, so keeping prompts clear, direct, and conversational will give you the best results.
- Ask the AI tool to show its working out: If you need AI to provide reasoning or logical steps, ask it to show its thought process. This technique, known as chain-of-thought prompting, is useful for tasks like explaining how it’s worked our a problem or outlining an argument.
- Refine as you go: Even with well-crafted prompts, the first output may not be perfect. AI works best if you refine – if a response is too vague or off-track, rethink your prompt by specifying the tone, level of detail, or word choice. Small adjustments can make a big difference.
AI tools can be incredibly useful, but they’re far from perfect. Regardless of which AI model you use, the quality of its output depends on the clarity and structure of your input.
Whether you’re using AI for brainstorming, automating tasks, creating a workout plan, or proofreading an article, thoughtful prompting leads to better, more accurate results.