What Is a Prompt?

Imagine you sit down at a piano and push a single key. A sound comes out — but it is just one note. Now imagine you play a carefully chosen chord, with the right timing and the right dynamics. Suddenly you have music. The piano had the same capabilities both times; what changed was the input. A prompt is your input to a generative AI model. It is the instruction, question, context, or example — or any combination of these — that you give the model before it generates a response. Everything the model produces is shaped by the prompt. A vague prompt is like pressing one key. A thoughtful prompt is like playing a chord: richer, more intentional, more useful. This lesson establishes what a prompt actually is; the next two lessons will teach you to write them well.

The Anatomy of a Prompt

A prompt can be as short as one word — 'Summarize' — or as long as many paragraphs of structured instructions. But prompts tend to contain some combination of five ingredients, even if not all five are always present. Task: What do you want the model to do? ('Write,' 'Explain,' 'List,' 'Compare,' 'Translate.') Context: What background information does the model need? ('I am a seventh-grader studying for a test.' 'This is for a school newspaper article.') Format: How should the output be structured? ('In three bullet points.' 'As a dialogue.' 'In exactly two sentences.') Constraints: What should the model avoid or include? ('No jargon.' 'Use only facts, not opinions.' 'Keep it under 100 words.') Examples: Show the model what good output looks like. ('Here is an example of the style I want: ...') Beginner prompts often include only the Task. Expert prompts layer in Context, Format, Constraints, and Examples — and the results are dramatically better.

Here is a concrete comparison. Consider these two prompts sent to the same language model: Prompt A: 'Tell me about photosynthesis.' Prompt B: 'Explain photosynthesis to a curious 12-year-old who has already learned that plants make their own food but does not know the chemistry. Use an analogy involving a kitchen. Keep your explanation to three short paragraphs. End with one question that would make a good experiment.' Prompt A will produce a generic, textbook-style explanation — possibly too advanced, possibly too basic, definitely not tailored to anyone. Prompt B uses all five ingredients: Task (explain), Context (12-year-old, what they already know), Format (three paragraphs), Constraints (kitchen analogy, no raw chemistry), and implied Example through the analogy instruction. The model is the same. The difference in output is entirely due to the quality of the steering.

System Prompts vs. User Prompts

When you use an AI chatbot or app, there are actually two kinds of prompts at work, even if you only see one of them. A system prompt is written by the developer who built the app. It sets the model's persona, rules, and context before you ever type anything. It might say: 'You are a friendly math tutor for middle school students. Always encourage students before giving the answer. Never solve the problem for them; guide them step by step.' A user prompt is what you type. The model sees both the system prompt and your message, and generates a response that honors both. When you use a consumer AI app, you are writing user prompts on top of a hidden system prompt you cannot see. This is why the same underlying model can feel very different in different apps — the system prompt shapes its behavior, tone, and capabilities.