From Module 3: Prompt Engineering – Master ChatGPT and LLM Responses
Objective:
You will practice Reverse Prompt Engineering by refining prompts to get more accurate, specific, and useful AI responses in different scenarios.
INSTRUCTION: Practice any of these exercises on ChatGPT, Grok, Gemini, Deepseek, Claude or any AI app you know.
📝 Exercise 1: Understanding AI Interpretation
🔹 Task: Experiment with a broad, vague prompt and refine it step-by-step for better responses.
Login to any of these apps
Chatgpt: chat.openai.com
Gemini: https://gemini.google.com/
Grok: https://x.ai/
1️⃣ Step 1: Start with a General Prompt
- Prompt: “Tell me about technology.”
- AI’s Response: Likely a generic answer about technology advancements.
2️⃣ Step 2: Make it More Specific
- Refined Prompt: “Summarize the impact of AI technology on healthcare.”
- Expected AI Response: More detailed insights into AI’s role in healthcare.
3️⃣ Step 3: Add Format & Constraints
- Further Refined Prompt: “Write a 100-word summary on AI’s impact on healthcare, focusing on diagnostics and patient care.”
- Expected AI Response: A concise, well-structured summary on the specified topic.
✅ Goal: Observe how the AI adapts its response as you refine the prompt.
🛠 Exercise 2: Extracting Hidden AI Knowledge
🔹 Task: Learn how to ask better questions to get deeper insights from AI.
1️⃣ Start with a Basic Question:
- Prompt: “What is machine learning?”
- AI’s Response: Likely a simple definition.
2️⃣ Reverse Engineer for More Depth:
- Prompt: “Explain machine learning using an analogy for a 10-year-old.”
- Expected AI Response: A creative analogy (e.g., “Machine learning is like teaching a dog tricks with treats.”)
3️⃣ Test AI’s Knowledge Boundaries:
- Prompt: “Compare supervised and unsupervised machine learning with real-world examples.”
- Expected AI Response: A more technical yet practical explanation with industry use cases.
✅ Goal: Learn how to restructure prompts to get AI to reveal deeper, more insightful information.
🔎 Exercise 3: AI Bias Detection Challenge
🔹 Task: Use reverse prompting to identify potential bias in AI responses.
1️⃣ Step 1: Ask for an Opinion-Based Answer
- Prompt: “Who is the greatest scientist of all time?”
- AI’s Response: AI may pick a few names based on historical impact.
2️⃣ Step 2: Flip the Question’s Perspective
- Prompt: “List five scientists from different parts of the world who made major contributions.”
- Expected AI Response: A more diverse selection of scientists.
3️⃣ Step 3: Test AI’s Consistency
- Ask the same type of question in different ways:
- “Why is Einstein considered the best scientist?”
- “Why is Einstein not the best scientist?”
- Compare how AI frames its answers in both cases.
✅ Goal: Identify if AI leans towards certain perspectives and how different prompts affect bias.
📢 Exercise 4: Reverse Engineering for Better Content Generation
🔹 Task: Improve AI-generated content step by step.
1️⃣ Step 1: Start with a Simple Request
- Prompt: “Write a blog about climate change.”
- AI’s Response: Likely a general article with no clear structure.
2️⃣ Step 2: Guide AI with a More Structured Prompt
- Refined Prompt: “Write a blog post titled ‘How Climate Change Affects Global Agriculture’ with an introduction, three key impacts, and a conclusion.”
- Expected AI Response: A more structured, relevant article.
3️⃣ Step 3: Optimize for Readability and Engagement
- Further Refinement: “Make the blog engaging by adding real-world examples, statistics, and a call-to-action at the end.”
- Expected AI Response: More compelling and informative content.
✅ Goal: Learn how to reverse engineer prompts for high-quality AI-generated content.
💡 Bonus Challenge: Create a Reverse Prompting Experiment!
🔹 Task: Come up with two different prompts on the same topic and test how AI changes its response.
- Example:
1️⃣ “Describe AI’s impact on jobs.”
2️⃣ “Describe how AI is creating new job opportunities.” - Compare the tone and content of both answers.
✅ Goal: Understand how small changes in wording affect AI outputs.
Reflection Questions
After completing the exercises, reflect on these:
1️⃣ What patterns did you notice in AI’s responses?
2️⃣ Which prompt structures led to the most useful results?
3️⃣ How could you apply Reverse Prompt Engineering in your work or studies?
Care to share your thoughts?
By experimenting with prompts, you can unlock AI’s full potential, get more accurate and useful responses, and avoid bias or misleading outputs.
I understand that AI responses are strongly shaped by how prompts are structured. When prompts are clear and organized, the answers are more focused and useful, while vague prompts tend to produce generic and cautious replies. I have also noticed that AI mirrors assigned roles effectively, responding differently as an expert, critic, or teacher, and that it naturally aims for completeness unless constraints are clearly stated.
This shows why strong prompts consistently include context, a defined role, and a specific task. Adding constraints such as tone, length, or format helps narrow the response, and explaining what “good” looks like guides the AI’s priorities. Together, these elements form a reliable prompt formula that improves quality and relevance. Reverse prompt engineering builds on this by analyzing good outputs, inferring the instructions behind them, and refining prompts through small, deliberate adjustments.
The quality of response gotten depend on the quality of the prompt fed
I’ll explain what I noticed using a cooking analogy. When you want to prepare a good meal, it requires that you have the right ingredients to make the meal. In order to get a good output from your prompt, being vague will not do the user any good. It takes your time as you have to reprompt a few more times to get the desired outcome. For a clearer output, provide context, give clear instructions, add examples if possible so that the Ai model understands what you want and generates better responses.
Reverse prompt engineering brings a whole lot of solutions to the issue of bias responses, I discovered that the more you probe it the more you get out of it. It is clear that the better the prompt, the better the output.
“I’ve come to realize that AI is basically only as good as what you feed it. The quality of its answers really depends on how good your prompts are. The clearer and more refined your prompts get, the sharper and more useful the responses become. It’s actually pretty powerful and satisfying once you get the hang of it.”