How to Write Better Prompts for AI

Start with the outcome, not the tool

A strong prompt begins with the result you want: a decision, plan, draft, review, clearer thinking, or next action. Instead of asking a broad question, define the task, audience, constraints, format, and quality standard.

Give the assistant a working role

Generic prompts often produce generic answers. Give the assistant a role that fits the job: strategy reviewer, writing editor, planning partner, customer-support analyst, or technical explainer.

Add constraints and red lines

Good prompts say what the AI should avoid: long lists, unsupported claims, jargon, vague advice, or skipping risks. These red lines are especially useful for work contexts where quality and decision discipline matter.

Use examples to calibrate quality

If you already know what a useful answer looks like, include a short example. Examples help the model understand structure, tone, and depth faster than abstract instructions.

Method and references

INSPIRE & CRAFTS research paper · OpenAI prompt engineering guide · Anthropic Claude prompt engineering docs · Google Vertex AI prompt design strategies

FAQ

What makes a prompt better?

A better prompt gives the AI a clear outcome, role, context, constraints, format, and quality standard. The goal is to reduce guessing.

Do I need a different prompt for every AI tool?

The exact wording may change, but the same core instructions can usually work across ChatGPT, Claude, Gemini, and similar assistants.