What is Prompt Engineering

Definition

Prompt engineering is the practice of designing and refining the inputs given to a language model to elicit accurate, relevant, and reliable outputs, using techniques such as clear instructions, examples, and structure to steer the model's behaviour without changing the model itself.

« Back to Glossary Index

Improves output quality without retraining or changing the model
Steers tone, format, and reasoning through carefully structured instructions
Reduces errors and ambiguity by providing examples and constraints
Unlocks complex tasks via techniques like step-by-step and few-shot prompting

Real World Example

An analyst improves an LLM's data-extraction accuracy by rewriting the prompt to specify the exact output JSON schema and including two worked examples, turning inconsistent responses into reliable, parseable results.

FAQs

What techniques does prompt engineering use?

Clear instructions, examples (few-shot), output-format specification, and step-by-step reasoning prompts are common techniques.

Why does prompt wording matter so much?

Models respond to context and phrasing, so precise, well-structured prompts substantially change the quality and reliability of outputs.

Is prompt engineering a substitute for fine-tuning?

Often yes for many tasks, since good prompting adapts behaviour without the cost of retraining, though fine-tuning suits deeper customisation.