How to write a good AI prompt

1. Give the model a role and a goal

Open with who the model should be and what you want. "You are a senior financial analyst. Help me understand this company's unit economics." A clear role and goal focus the response far more than a vague question.

2. Provide the context it needs

Models cannot read your mind or your files. Paste the relevant text, data, or background. State the audience, the format you want, and any facts that matter. Most weak responses come from missing context, not a weak model.

3. Be specific about the output

Say exactly what you want back: a numbered list, a table, a 200-word summary, code in a specific language. If structure matters, describe it. Ambiguous asks produce ambiguous answers.

4. Show an example when you can

One good example of the input and the output you expect (a "few-shot" example) often beats paragraphs of instructions. Show the model the shape of a great answer.

5. Set constraints and guardrails

Tell it what to avoid, how long to be, what tone to use, and what to do when unsure. "If you are not certain, say so rather than guessing" is one of the highest-value lines you can add.

6. Iterate

Treat the first answer as a draft. Point at what is wrong and ask for a revision: "Tighten this, drop the marketing language, and add a concrete example." Prompting is a conversation, not a single shot.

Want ready-made examples? Browse the prompt library: every prompt already follows these habits and tells you the context to provide.