🤔 How much does your prompt affect what AI gives you? Let’s put it to the test.
Same documents.
Same LLM.
Two different prompts.
You don’t need technical language or a secret formula. In this exercise, you’ll see what changes when you tell AI exactly what you need—and how you plan to use it.
👇 Try both rounds, then judge the results for yourself.
1️⃣ Prompt: Keep it simple
Open a new chat in your favorite LLM, upload Documents A and B (attached), and type just one word: "Compare"
Review the response.
2️⃣ Prompt: Give clear, concise directions
Open another new chat in the same LLM. Upload both documents again, then copy and paste this prompt:
“Compare Document A (original) with Document B (revised). Create a professional, clearly formatted Word comparison document that I can download and share with my boss. Include a brief executive summary followed by a comparison table. For each difference, identify the section, quote the exact old and new text, label it added, removed, or changed, and explain the effect on meaning. Use only the uploaded sources. Flag uncertainty instead of guessing.”
Now compare the two results.
Which caught more differences?
Which was easier to verify?
Which gave you something you could actually use?
Try switching model and/or thinking settings up and down?
Think of it like assigning work to a coworker. How much does the result depend on how clearly you explain the task? Share your observations in the comments.