The information, sentences, or questions that you enter into a Generative AI tool (“prompts”) are a big influence on the quality of outputs you receive. After you enter a prompt, the AI model analyzes your input and generates a response based on the patterns it has learned through its training. More descriptive prompts can improve the quality of the outputs.
This guide gives you a basic overview of how to Generate better prompts. There are more resources at the end, and we’ll be updating our guidance over time as we learn how generative AI is being used at the University.
It’s very important to keep in mind that AI-generated content can be inaccurate, misleading, entirely fabricated, or offensive, so be sure to carefully review any work containing AI content before you use or publish it.
- Before you start using an AI tool for your Harvard work, you should be sure to review the University’s guidelines.
-
Harvard's Office of the Vice Provost for Advances in Learning (VPAL) has also produced a comprehensive collection of customizable prompts tailored specifically for educational use.
Be specific
Generic prompts like “Write a story” will produce generic results. What kind of story do you want? What genre? Is it for adults or children? How long should it be? Is it funny or serious? Adding this type of context and being specific, clear, and concise will help generate more useful outputs.
More specificity can also limit the chances of inaccurate responses. If you’re looking for a good meal near Harvard Square, asking “What is the best restaurant in Cambridge?” could return results from Cambridge, UK. Asking “What is the best restaurant in Cambridge, Massachusetts, within walking distance of Harvard Yard” gives you a better chance of finding what you’re looking for.
“Act as if…”
Asking the AI to behave as if were a type of person, process, or object can be an easy way to start generating better prompts. The AI will attempt to emulate that role and tailor its answers accordingly.
For example, imagine you wanted help creating a recipe based on ingredients you have in your kitchen. If you ask, “create a recipe using these ingredients,” followed by a list, you’ll get some general recipe suggestions. But if you added “act as if you are my personal trainer” first, the AI will consider this context in its response, perhaps suggesting a healthier recipe or a meal designed to refuel after a workout.
Tell it how you want your output to be presented
Generative AI can produce many different types of outputs, including code, stories, reports, summaries and synopses, dialogue, business communications, audio, images, and much more. Being specific about the type of output you want will produce better results. After describing what you want, try adding “Present this in the form of…” and your preferred output. Or start your prompt with, “Create a [format of output] about / that contains…” etc.
Use “do” and “don’t”
Telling AI what you do and don’t want in your response can save time and improve your result. To use the recipe example again, you can specify what foods you do want to include in your meal, and any that you specifically do not want, e.g. “Act as if you are my personal trainer. Create a recipe that will help me refuel after my workout. Do include tomatoes, chicken, and some type of carb. Don’t include chili peppers or any ingredients containing wheat.”
Use examples
Using an example of a sentence or paragraph that you’d like your output to resemble can give the AI a reference point to help improve results. However, it’s very important to not infringe on copyright. Don’t upload copyrighted works and ask the AI to rewrite / reproduce it or create a substantially similar work.
Consider tone and audience
Give the AI specifics about who your audience is and what sort of tone you’d like to set. For example, “Give me ideas for a best man’s speech that is funny and heartwarming but appropriate for a family audience” will generate better results than just “Write a best man’s speech”. Adding more specificity and context will further improve the output.
Build on previous prompts
You don’t have to get everything into your first prompt. Try starting with a basic question and adding to it over time. Change the wording or tone or add more context and specificity to guide the AI toward the output you’re looking for.
Correct mistakes and give feedback
Chat with the AI as if it’s a colleague or teammate and you’re working on a project together. Give feedback – tell it which parts of the output were useful and which parts could be improved. If you notice it got something wrong, tell it so it can correct its mistake.
Ask it to create your prompts or what else it needs from you
Stuck on how to create a good prompt? Ask AI to help by creating one for you! Start with a basic idea of what you want and ask the AI to expand on it for you, like “What should I ask you to help me write a blog post about AI?”. And simply adding “Tell me what else you need to do this” at the end of any prompt can fill in any gaps that will help the AI produce better outputs.
More resources
Please note HUIT is not responsible for content on third-party websites:
- HUIT Generative AI website
- Examples of ChatGPT prompts, especially “Act as if…” prompts
- ChatGPT prompt engineering for developers
- University of Michigan: Prompt literacy in academics