Prompt Engineering Quiz: Which Prompt Gets the Better Answer?

The difference between a useless AI answer and a great one is usually the prompt, not the model. This prompt engineering quiz mixes two kinds of questions: pick the prompt that will get the most useful answer from four candidates, and answer concept questions about techniques such as few-shot examples, role prompting, chain-of-thought, format instructions, and iteration. Ten questions are drawn from a pool of twenty-four, and each explanation says why the winning prompt wins.

The principles hold across GPT, Claude, Gemini, Grok, and DeepSeek: give context, name the audience, specify the format, show an example, and verify what comes back. If you have ever typed 'make it better' and been disappointed, this is for you. Use it as prompt engineering practice questions for a team, or take it twice and see how many of the winning prompts you can write from memory.

Sample questions

1. You need an email to your landlord about a heater that has been broken for five days. Which prompt gets the most useful draft?
   • Write an email to my landlord about the heater.
   • My heater is broken, help.
   • Write a polite but firm email to my landlord: heater broken 5 days, I've called twice, I need a repair date this week. Under 150 words.
   • Write a long, formal, legal-sounding letter to my landlord.
2. You want to explain compound interest to a 12-year-old. Which prompt gets the most useful answer?
   • Explain compound interest to a 12-year-old using a weekly allowance example, in under 100 words, then ask one question to check they got it.
   • Explain compound interest.
   • Give me the formula for compound interest.
   • Write an essay about interest rates for kids.
3. You need to summarize a long article for your manager. Which prompt gets the most useful result?
   • Summarize this.
   • What is this article about? [paste article]
   • Make this article shorter: [paste article]
   • Summarize the article below in five bullets for a busy manager, then list any numbers or claims I should double-check: [paste article]

Start the quiz above to see the answers and explanations. Each attempt draws 10 questions from a pool of 24.

What good prompts have in common

Every winning prompt in this quiz does the same few things. It gives the model the material it needs, says who the answer is for, names the format, sets limits, and asks for something checkable. The pattern is the same whether you are drafting an email, debugging code, or planning a trip:

• Context: what happened, what you have, and what you already tried
• Audience: who will read the answer and how much they know
• Format: bullets, a table with named columns, a word limit, a tone
• Examples: two or three samples of the style you want
• Constraints: what to leave out, stated with the alternative you prefer
• Verification: ask for sources or a list of claims to double-check

Prompting habits that save time

Iterate in the same conversation instead of starting over; the model already has the context, so tell it what to keep and what to change. For anything complicated, invite clarifying questions before it answers. Split large jobs into steps and review between them, because a single sprawling request spreads the model's attention thin and lets early mistakes propagate.

Two things prompts cannot do: change the temperature setting, which lives in the app or API, and make the model know things it was never given. Paste the source text, check any statistics against the original, and use a system prompt for rules you would otherwise repeat. Switching models on a hard question is a legitimate move too; in ChatUp the conversation comes with you.