Zero-Shot Prompting

The plain-English definition

Zero-shot prompting means giving the AI a task with zero examples. You describe what you want and let the model figure out how to do it. This is what 99% of people do every time they open ChatGPT.

Zero-shot example

"Translate this English text to French: I love coffee in the morning."

You haven't shown the model any English-French pairs. It just translates because it knows the task.

When zero-shot works well

• Simple, well-defined tasks (translate, summarize, define)
• Anything the model has seen variations of in training
• Quick exploration where you don't know what good output looks like yet

When zero-shot fails

• Tasks with a specific format you want copied (use few-shot instead)
• Highly domain-specific work where the AI lacks training
• When tone matters and you want a consistent voice (show 2 examples first)

Zero-shot vs few-shot, in one line

Zero-shot says: "Do this task."
Few-shot says: "Here are 2 examples. Do this task."

Few-shot wins on consistency. Zero-shot wins on speed.

Related terms

• Few-Shot Prompting
• Chain-of-Thought
• Role Prompting
• Prompt Engineering