A prompt is the instruction and context handed to a model at inference time. Because large models perform in-context learning (Brown et al., 2020), the prompt — not retraining — is the primary way most users steer behaviour. The systematic study of prompting methods is surveyed by Liu et al. (2023).
Effective prompts state the task clearly, supply relevant context or examples, and specify the desired format. Small wording changes can meaningfully shift results — which is exactly why prompt engineering exists.
What is prompt engineering?
Prompt engineering is the skill of writing, testing and iterating on prompts deliberately rather than by trial and error. Instead of changing the model, you change its input: you make the task explicit, give the model the background it needs, show what a good answer looks like, and say how the result should be structured.
It applies to everyday chat as much as to developers building on an API. For applications, prompt engineering also covers the system prompt that sets a model's role and rules, and defending against prompt injection. OpenAI, Anthropic and Google each publish prompting guides for their own models (see the references below).
Prompt engineering techniques that work
The same request, written vaguely and then with each technique applied:
Prompt engineering at MultipleChat
MultipleChat's Prompt Optimizer improves a rough prompt before you send it. Because the same prompt can run on ChatGPT, Claude, Gemini, Grok and Perplexity side by side, you can also see how a wording change shifts each model's answer.