Imidef Blog

Complete Guide to Writing Prompts｜Tips and Specific Examples of Using ChatGPT

Prompt writing illustration

If ChatGPT keeps giving you vague answers, the problem is usually not the model—it is the prompt. Prompts are the instructions that set context, clarify expectations, and steer the AI toward the result you want. This guide walks through a proven structure for writing high-performing prompts, complete with examples you can adapt to your projects.

What Exactly Is a Prompt?

A prompt is any text you send to the AI to initiate a task. It can be a question, a directive, or even a conversation starter, but the principle is the same: the clearer your instructions, the better the output.

Three-Step Framework for Prompt Writing

  1. Declare the goal upfront.
    • "Translate this email into natural English."
    • "Draft a Python script that fetches weather data." Leading with the objective helps the model lock onto the intent of your request.
  2. Layer in constraints and context.
    • Specify audience, tone, length, or tools: "Use a friendly voice," "Limit to 200 words," "Reference recent research." Constraints act like guardrails, reducing the chance of irrelevant tangents.
  3. Describe the desired format.
    • Markdown outline, JSON, table, checklist—spell it out.
    • Provide examples if the format is unusual. When the model knows how the answer should look, the result is immediately usable.

Fixing Common Prompt Mistakes

  • Too short: "Write a blog post."
    • Result: generic fluff.
    • Fix: add topic, audience, tone, length, and structure.
      • "Write a 700-word blog post about digital detox for beginners. Include an introduction, three H2 sections, and a summary with actionable tips."
  • Missing context: "Explain this code."
    • Fix: paste the code, describe what confuses you, and share the runtime environment.
  • Conflicting instructions: "Keep it concise" plus "Include every detail."
    • Fix: prioritize what matters most or split the prompt into stages.

Advanced Technique: Assign a Role

Role-based prompts shape the model's perspective. By framing the AI as a specific expert, you nudge it to use relevant jargon, processes, and frameworks.

  • "You are a veteran product manager. Critique this roadmap and highlight three risks."
  • "You are a front-end engineer. Provide a Next.js component that matches the following design specs."

Combine roles with the three-step framework for maximum clarity.

Bonus Tips for ChatGPT Prompt Design

  • Stack prompts: Break large tasks into a series of smaller prompts and feed the AI its previous answers.
  • Request alternatives: Ask for multiple versions to increase creative range.
  • Iterate out loud: Share what you liked or disliked in the last answer so the AI can refine the next one.
  • Save templates: Keep a library of your best prompts for repeated tasks like summaries, reports, or code reviews.

Summary

  • Strong prompts clarify goal, context, and output format.
  • Examples and constraints reduce ambiguity and improve accuracy.
  • Role assignments and iterative feedback unlock expert-level responses.

Want to go deeper? Pair these tactics with GPT-5's thinking modes in your next session to see how response quality evolves.

Related articles