What is a prompt and what does it do?
A prompt is an instruction to the AI or Large Language Model (LLM). Since they’re written in natural language, they are straightforward to create and edit. But writing reusable or shareable prompts can be more complicated. Our prompt library and technical reports can help you discover, refine, build, and test your prompts and provides examples of prompting approaches.
Spotlight: Sample Prompt
“Devil’s Advocate”
This prompt is designed to generate ideas to solve problems. It is based on ideation principles including generating many ideas, using constraints, and integrating perspectives.
How to Use Prompts
Prompting with LLMs always begins with a goal in mind. What do you want to accomplish? What does success look like? Coming to the AI with a prepared mind is an important part of developing a reusable prompt.
Discover
Our prompt library is the perfect place to begin exploring and discovering what's possible.
Analyze and Test
It’s important to test prompts that you will reuse or share with others, taking into account a variety of user perspectives.
Create and Refine
Make our prompts your own by adapting them to your specific context.
The Library
Our prompt library contains carefully designed and evidence-based prompt templates organized by purpose. Each prompt includes clear instructions, suggested use cases, and model options.
Prompts on this page (but no other content on the site) are licensed under Creative Commons License Attribution 4.0 International This license requires that reusers give credit to the creators (Ethan Mollick and Lilach Mollick). It allows reusers to distribute, remix, adapt, and build upon the material in any medium or format, even for commercial purposes. Use prompts at your own risk, outputs may not be correct.
Analyzing and Testing Prompts
A prompt that works well in one model might produce different results in another, and even the same model can respond differently to the same prompt over time. Consider the following when working with a new prompt you want to share or an updated model:
- Test your prompt multiple times to ensure consistent results
- Consider how the prompt might work for others
- Check if the prompt stays on track during extended interactions
- Verify that the prompt provides the expected output
- Try the prompt with a variety of models to understand the variation in responses
- Test edge cases to see how the prompt handles unexpected inputs
- Ensure the prompt aligns with your goals
Creating and Refining Prompts
You can expand existing prompts with additional context, edit them to focus on particular goals, or use our prompts as a basis to create entirely new prompts from scratch. Here are some things to keep in mind as you do so:
- Start with clear goals
- Include context – information, persona’s, perspectives, explanations, documents, or files
- Provide step-by-step instructions for the AI to follow
- Add examples of good outputs
- Test and iterate based on actual interactions
Questions? Contact Us
