Prompt technique
Basic / Zero-Shot Prompting
The simplest, most direct form of prompting — just ask the model what you want.
What it is
Basic (zero-shot) prompting is the foundational technique where you give a model a clear instruction without providing examples. Modern frontier models like GPT-4o, Claude Opus and Gemini Pro perform remarkably well on zero-shot tasks when the instruction is specific, the role is clear, and the desired output format is stated. Use it for simple, well-known tasks.
When to use it
- ✓Simple, well-defined tasks the model has clearly seen during training
- ✓Quick drafts, summaries, translations, definitions, or rewrites
- ✓When you want to benchmark a model's raw capability before adding scaffolding
Example
Summarize the following article in 3 bullet points, written for a non-technical audience.
{article}Why it works: A clear instruction, an explicit format (3 bullets), and an audience constraint — no examples needed.
Pitfalls
- !Vague instructions produce vague outputs — be specific.
- !Zero-shot fails on novel domain-specific formats; reach for few-shot instead.