The Five-Part Prompt: Role, Context, Examples, Constraints, Format

Quick prompts work for quick tasks. But when you want something solid — a first draft of an essay, a piece of code, a study plan — you need structure. Most professional prompt engineers use some version of this five-part recipe.

Putting it all together

You are a patient middle-school math tutor who loves real-world analogies.

I'm a 7th grader studying for a test on linear equations. I understand what a variable is but I struggle with word problems.

Here's the kind of hint I find helpful:
> 'Think of x as how many apples are in the bag — your job is to figure out what number makes the sentence true.'

Constraints:
- No solving the problem for me.
- Give one hint at a time.
- Use analogies, not algebra vocabulary.

Format your response as:
**Hint:** <one sentence>
**Try this:** <one guiding question>

Here's my problem: A bookstore sells paperbacks for $8 and hardcovers for $15. They sold 12 books total and made $128. How many of each did they sell?

Notice how every part has a job. The role puts the AI in tutor mode. The context tells it what you know. The example models the tone you want. The constraints stop it from spoiling the answer. The format makes the output scannable. You will get a wildly better response than 'help me with this math problem.'

Why role matters so much

When you tell the AI 'you are an experienced editor' or 'you are a skeptical scientist,' it shifts its whole voice. It pulls on different patterns from its training. This one line often changes the answer more than any other part of the prompt.

• 'You are a careful research assistant who cites sources.'
• 'You are a senior software engineer reviewing a junior's code.'
• 'You are a Socratic teacher who only asks questions back.'
• 'You are a stand-up comedian writing jokes for a school talent show.'