Plain definition: Zero-shot prompting is when you ask an AI to do a task without giving it any examples—you simply describe what you want and trust the AI to figure it out from its training alone.

In plain terms

It’s asking someone to do a task cold, with no sample to reference. “Write me a one-paragraph bio for this executive” is a zero-shot prompt. You haven’t shown the AI what your ideal bio looks like—you’re relying on its general understanding of what a bio is. This works well for common, well-understood tasks; for unusual or highly specific tasks, providing examples (few-shot prompting) gets better results.

Why it matters for operators

Most everyday AI use is zero-shot—you type a request and go. Knowing the term helps you recognize when zero-shot isn’t working well and you need to switch strategies. If you’re getting inconsistent or off-target results from a task, that’s often a signal to move from zero-shot to few-shot prompting by adding two or three examples of what “good” looks like.

Example

A marketing agency asks an AI to classify 200 customer testimonials as “feature-focused,” “service-focused,” or “value-focused.” They try zero-shot first—just describing the categories—and find the AI is inconsistent. They switch to few-shot, add three labeled examples per category, and accuracy jumps significantly across the whole batch.