Sometimes it's hard to describe what you want, but easy to show it. That's the idea behind few-shot prompting: you include a few examples of input and the ideal output, and the AI follows the pattern.
Zero-shot vs few-shot
A zero-shot prompt only gives instructions: “Write a product tagline.” A few-shot prompt adds examples:
Write a tagline in this style.
Product: noise-cancelling headphones
Tagline: Silence, on demand.
Product: reusable water bottle
Tagline: Every sip, a small promise.
Product: [your product]
Tagline:The examples communicate length, tone and rhythm far more precisely than adjectives ever could.
When to use it
- Matching a specific brand voice or writing style
- Consistent formats, such as product descriptions or support replies
- Classifying or tagging items the same way every time
- Turning messy data into a fixed structure
Tips for good examples
- Quality over quantity: two or three excellent examples beat ten average ones.
- Show variety: if all examples are about food, the AI may assume every answer should be too.
- Keep the format identical across examples, using the same labels and structure.
- Put your real input last, in the same format, so the AI knows exactly where to continue.
Combine with instructions
Few-shot works best alongside a short instruction explaining the goal. The instruction says what to do; the examples show how it should look.