What is a prompt?
Everything you type to an AI is a prompt. Here's how a few small changes get much better answers.
Every message you send to an AI is a prompt. “Prompt engineering” sounds technical, but it mostly means one thing: being clear about what you want.
The text you give an AI model: your question, your instructions, and any background it needs.
Briefing a new colleague. Say “write an email” and you get something generic. Explain who it’s for, why, and what tone you want, and you get something you can actually send.
The same request, two ways
What a good prompt usually includes
- 1Role — who the AI should act as, e.g. a support agent
- 2Context — the background it can't know on its own
- 3Task — exactly what you want done
- 4Format — length, tone, bullet points or paragraphs
- 5Example — optional, but shows what good looks like
Two kinds of prompts
- System prompt: instructions set behind the scenes by the app, like “You are a helpful support assistant. Be concise.” You usually don’t see it.
- User prompt: what you type in the chat box.
A common misunderstanding
“There are magic words.” Clever tricks matter far less than clear instructions and good context. If a person would be confused by your request, the AI probably will be too.
A recruiter who asks for “a job post for a developer” gets something generic. Adding the role, the level, the location, the salary range and the tone the company uses gets a draft they can post with small edits.
Related buzzwords
LLM · Context window · Temperature · Hallucination · Prompt injection