ChatGPT Prompting

Introduction

ChatGPT (GPT-4o, GPT-5) responds best to clear instructions, markdown formatting, and JSON mode for structured output. It's also the most popular LLM, so prompts here ship to billions.

Beginner analogy: ChatGPT is the friendly generalist — clear instructions go a long way; it's chatty by default.

Understanding the topic

Core concepts to understand:

• Use the system message for persistent instructions.
• Enable JSON mode when you need structured output.
• Use tools (function calling) for actions.
• Temperature 0–0.3 for factual; 0.7+ for creative.
• Markdown renders beautifully in ChatGPT UI.

Syntax reference

Visual workflow / architecture:

messages = [
  { role: "system", content: "You are a senior dev reviewer." },
  { role: "user",   content: "Review this PR: ..." }
]
response_format = { type: "json_object" }

Real-world use

OpenAI's own Custom GPTs are just elaborate system prompts. 5M+ have been published.

Best practices

• Front-load critical instructions; ChatGPT respects early system content.
• Use JSON mode when downstream parses output.
• Keep system < 500 tokens for cost.

Hands-on exercise

Interview preparation — practice these questions:

• Q1. How get reliable JSON from ChatGPT?
• Q2. System vs user — what goes where?
• Q3. Function calling in 1 sentence?
• Q4. Temperature defaults?