If I asked a freelance writer for “a post about our new feature, make it engaging”, I would expect one of two replies. A good one would send back four questions. A bad one would send back a draft I did not want, and politely invoice me for it.
We all know this about people. We forget it entirely the moment the writer is a model, and then we blame the model for the missing brief.
A brief has four parts. None of them takes longer than a sentence.
1. The reader
Name the person. Not a demographic, a person in a situation. “A studio owner who has just been asked by a client why last month’s campaign cost what it did.”
The reason this comes first is that it silently decides everything below it. Vocabulary, what you can assume, what has to be explained, how much patience you have earned before the first useful line.
Almost every flat draft I have received, human or machine, traces back to a missing or generic reader. “Small business owners” is not a reader. It is a market.
2. The one job
What should be different after someone reads this? One thing.
“They understand why we changed the pricing.” “They book a call.” “They stop believing the most common myth in our category.” Pick one and say it plainly, because a piece asked to do three jobs does none of them and takes three times as long to write.
This is also the part that makes revision possible. If the job is written down, “this does not do the job” is a specific, checkable criticism. If it is not, all you can say is that the draft feels wrong, and feels wrong is not a note anyone can act on.
3. The constraints
Length, format, what must appear, what must not. The banned list matters more than the required list.
“Under two hundred words. No question as the opening line. Mention the migration, do not mention the competitor. British spelling. No exclamation marks.”
These are boring to write and they are where most of the rework disappears. A constraint you did not state is a constraint you will enforce later, one draft at a time, by hand.
4. One reference
Show one thing that is close to right and say what is right about it.
Not five references. One, with a sentence: “This, because the second paragraph gives a number before it gives an opinion.” A single annotated example carries more instruction than a paragraph of adjectives, which is true of new writers on your team as well.
If you cannot find a reference, that is information. It usually means you have not decided what you want, and no amount of briefing will rescue the draft.
Why “make it engaging” fails
“Engaging” describes how you want to feel about the result, which makes it a review rather than an instruction. It says nothing about how to get there, which tells the writer nothing about how to get there.
A freelancer will absorb one round of this and start guessing, because guessing is faster than another email. A model will absorb an unlimited number of rounds, cheerfully, and produce a slightly different guess every time. That is the trap: the cost of a bad brief is invisible, so it never gets fixed.
”Engaging” is a review of the result, and a review tells the writer nothing about how to get there.
Argue with the brief, not the draft
Here is the habit that changed my hit rate more than any prompt technique.
When a draft comes back wrong, do not reply to the draft. Do not write “more punchy, less corporate” and try again. Go back to the four lines, find the one that was vague, fix it, and rerun.
Nine times out of ten the fault is visible the moment you look. The reader was a market. The one job was two jobs. The constraint you are now complaining about was never written down.
Editing the brief compounds, because next time you start from a better one. Editing the draft does not compound at all, and it is how people end up on version nine of something they should have thrown away at version two.
Keep the briefs that worked
When a brief produces something you publish, save it. Four lines in a document, with the name of the piece it produced.
After a couple of months you will have eight or ten, and you will notice they fall into shapes: the customer story brief, the feature announcement brief, the myth-correction brief. Then the work becomes filling in a shape rather than starting from a blank page, which is the actual promise of working this way.
This is also the most transferable asset in the whole exercise. Prompts go stale as tools change. A brief that describes your reader, your job and your constraints is still correct in two years.
The half you only write once
Two of the four parts change with every piece: the one job and the reference. Two barely change at all.
Your reader is mostly stable. Your constraints are mostly your brand voice: the words you avoid, the moves you do not make, the register you hold. Write those down once, properly, and half of every brief is already finished. I have argued elsewhere that five lines are enough for the voice half, and the same document doubles as the standing part of every brief you write afterwards.
That is why briefing well and having a voice are the same project. A voice document nobody reads is decoration. A voice document pasted into the top of every brief is infrastructure.
Where to start
Take the next thing you were going to ask a model for and write the four lines first, on paper, before you open the tool. Reader, one job, constraints, one reference.
Give yourself five minutes for it. If the draft that comes back is still wrong, you now have something to fix that is four sentences long instead of eight hundred words.
