Types of Prompts
This page describes common types of prompts used with AI language models. Each type includes a definition, example, and when to use it.
1. Zero-shot Prompting
What it is: You give the model a task without examples.
Example
Q: Translate this sentence into French: I love pizza.
A: J'adore la pizza.
When to use: Simple or well-known tasks.
2. Few-shot Prompting
What it is: You show the model a few examples of how you want it to respond.
Example
Q:
Translate into French:
– I love pizza. → J'adore la pizza.
– She is happy. → Elle est heureuse.
– He is tired. →
A: Il est fatigué.
When to use: To guide the model with patterns.
3. Chain-of-thought Prompting
What it is: You explicitly ask the model to reason step-by-step.
Example
Q: If a train leaves the station at 3 PM and travels for 2 hours, what time does it arrive? Let's think step by step.
A: The train leaves at 3 PM. It travels for 2 hours. 3 PM plus 2 hours is 5 PM. So, the train arrives at 5 PM.
When to use: Math, logic, and complex reasoning.
4. Instruction-based Prompting
What it is: You give very clear, structured instructions.
Example
Q: Summarize this paragraph in 2 sentences focusing only on the main idea.
A: The paragraph discusses the importance of teamwork in achieving goals. It emphasizes that collaboration leads to better results than working alone.
When to use: For high precision or consistent format.
5. Role-based Prompting
What it is: You assign a role or persona to the model.
Example
Q: You are a professional lawyer. Explain this legal term in simple language: "Habeas Corpus".
A: "Habeas Corpus" is a legal term that means a person has the right to be brought before a judge if they are being held in jail, to make sure their detention is lawful.
When to use: To match tone, expertise, or audience.
6. Contextual or Dynamic Prompting
What it is: You insert relevant background/context before the question.
Example
Q: In the context of European history during the 19th century, explain the causes of the French Revolution.
A: The French Revolution was caused by social inequality, economic hardship, and the influence of Enlightenment ideas, which led people to demand more rights and a fairer government.
When to use: When the task depends heavily on prior information.
7. Multi-turn Prompting
What it is: Building context across a conversation.
Example
Q: (Turn 1) Imagine you are a travel agent. I want to visit Italy.
A: (Turn 1) That sounds wonderful! What cities in Italy are you interested in visiting?
Q: (Turn 2) I'm interested in Rome and Venice. What are the must-see attractions?
A: (Turn 2) In Rome, you should visit the Colosseum, the Vatican, and the Trevi Fountain. In Venice, don't miss St. Mark's Basilica, the Grand Canal, and a gondola ride.
When to use: For dialogue, memory-based reasoning, tutoring, etc.
8. Prompt Chaining (Modular Prompting)
What it is: Breaking tasks into smaller steps and using the output of one as the input of the next.
Example
Q (Step 1): Extract all dates from this text: "The conference was held on March 3, 2022, and the next meeting is scheduled for July 15, 2023."
A (Step 1): March 3, 2022; July 15, 2023
Q (Step 2): For each date, summarize the event that happened.
A (Step 2):
- March 3, 2022: The conference was held.
- July 15, 2023: The next meeting is scheduled.
When to use: For complex tasks like code generation, document parsing, etc.