One-Shot Prompting
Learning to Fold a Saree by Watching Once
4 min read
Learning to Fold a Saree by Watching Once
Your grandmother shows you once how to fold a saree perfectly — now you can do the rest yourself.
One example is enough to establish the pattern. You don't need 20 demonstrations — just one clear example anchors the format in your mind. One-shot prompting works the same way: give the AI one example of exactly what you want, and it replicates the pattern for your actual input.
In Plain English
One-shot prompting gives the AI one example before your actual request. The single example demonstrates the format, style, or structure you want, and the AI follows that pattern.
The Technical Picture
One-shot prompting provides a single input-output demonstration pair before the target query. This in-context learning signal anchors the model to a specific format or task framing, often significantly improving consistency over zero-shot for structured output tasks.
Real-World Examples
- Showing one example of a product description format before generating 50 more
- Providing one sample customer reply before asking the AI to handle new complaints
- One example tweet before generating a content calendar
One example is often enough to lock in the format — one-shot is the sweet spot between zero and multi-shot.