Write prompts that get usable answers
Most disappointing AI answers come from a missing brief, not a weak model. A four-part structure fixes most of them.
- Level
- Beginner
- Reading time
- 8 minutes
- Sources checked
- 11 Nov 2026
- Status
- Draft, awaiting editorial review
Who it’s forAnyone who has tried an AI chat tool and found the answers generic.
Before you startNone. It helps to have read what an AI model actually does.
By the end you'll be able to
- Spot the four things a good prompt tells the model
- Rewrite a vague request into a clear brief
- Check an answer before you rely on it
Why prompts fail
A language model only knows what is in its training and what you put in front of it. When you ask for "a marketing email", it has to guess who it's for, what you're selling, how long it should be and what tone you want. It fills each gap with the most average answer, and average is rarely what you needed.
The fix isn't magic words. It's the same thing you'd give a capable new colleague: a short brief.
The four-part brief
Work through these in order. Each one removes a guess.
- 1GoalWhat you want to end up with, and what it's for. "A 120-word email inviting past customers to a spring open day."
- 2ContextThe facts the model can't know: your audience, your product, what's been tried, what must be included.
- 3ConstraintsLength, tone, reading level, things to avoid, and anything that must be exactly right, such as dates or prices.
- 4Output formatHow the answer should be laid out: a list, a table, three options to choose from, or plain text you can paste.
Sort it
Name the part
Each line comes from a brief for a café's new-hours handout. Which of the four parts is it?
Need a hint?
Ask of each line: is it what I want, a fact about my situation, a limit, or a layout?
A worked example
Before and after
Spot what the brief adds
Write a marketing email for my bakery.
Write a 120-word email inviting past customers of a small family bakery to our spring open day on Saturday 4 April, 10am to 2pm. Customers are mostly local families. Mention free tasting of three new sourdoughs. Warm and plain, no exclamation marks, no discount codes. Give me two subject-line options and the email body.
The second version is longer to write and far quicker to use. Every sentence in it answers a question the model would otherwise guess.
Iterate instead of starting over
- If the first answer is close, say what to change: "shorter", "less formal", "keep paragraph two, rewrite the rest".
- If you're not sure what to include, ask the model to ask you questions first: "Before you write, ask me up to five questions you need answered."
- Keep a prompt that works. Save it as a template and fill in the context next time.
Check the answer before you use it
A useful habit: ask the model which parts of its answer it is least sure about. It won't always know, but it often points you to the right places to check.
Try it
Pick something you asked an AI for this week that came back generic. Rewrite it with all four parts and compare the two answers side by side.
Your turn
Rewrite a vague request
Written answers aren’t marked automatically and aren’t saved: what you type stays on this page until you leave it. Compare yours with our example and the checklist instead.
Free worksheet
The four-part brief
Turn a vague request into a clear brief: goal, context, constraints and output format, with a filled-in example.
The online version saves your answers only in this browser. The PDF is A4 and prints on US Letter too.
Your progress
You’ve finished 0 of 3 activities in this lesson.
When you've tried the practice task on your own work, mark the lesson complete. It's your call: finishing the activities shows you've practised, not that you've mastered the skill.
Progress is saved only in this browser, on this device. There’s no account, and nothing is sent to us or anyone else. Clearing your browser data, or using another browser or device, starts you afresh. We save which activities you finished, never what you type.
Suggested next lesson
Check an AI answer before you rely on itA clear brief gets a better answer, not a guaranteed one. Next, a routine for checking an answer before you rely on it.Beginner, 12 minutesNext steps
Sources and evidence
A method guide. The brief structure is our editorial advice, not a measured result; the warning about confident errors is sourced below.
- OpenAI: Why language models hallucinate (5 Sep 2025) (opens in a new tab)Checked 11 Nov 2026
This guide is a draft and hasn’t been through editorial review yet. How we review guides