The industry decided the skill was prompting. It is not. The skill is briefing, and it is a skill most people in a marketing role already have and do not realise transfers.
An AI content brief is a short written instruction that gives the model context, audience, constraints, and a definition of good before asking for output. It is the same information you would give a competent freelancer. Briefing produces usable content because it supplies the specifics a model cannot know, whereas prompt formulas mostly change the shape of generic output.
Why prompt tricks disappoint
Prompt libraries full of clever phrasings work on the margin. They are not the difference between garbage and usable, because they do not address the actual problem.
The actual problem is that the model has no idea what your business knows. It has read an enormous amount of writing in your category and, absent instruction, it produces the statistical centre of all of it: fluent, competent, structurally correct, entirely interchangeable.
No arrangement of magic words extracts a specific from a model that was never given one. Specifics have to come from you, and a brief is the mechanism for handing them over.
The six components
| Component | Answers | Why it changes the output |
|---|---|---|
| Audience | Who reads this, in what situation? | Determines vocabulary, assumed knowledge, and what to skip |
| Purpose | What should they do or believe after? | Stops the model hedging toward general usefulness |
| What only you know | Your numbers, examples, and observations | The single biggest quality lever |
| Constraints | Length, format, banned words, claims you cannot make | Prevents the output you would have to rewrite |
| What good looks like | An example of your own writing | Shifts voice more reliably than adjectives |
| Structure | Sections, and what each one does | Stops the model organising it its own way |
The third row is where nearly all the quality lives, and it is the one people skip because it requires thinking rather than typing.
A template you can reuse
AUDIENCE
Who this is for, their situation, what they already know.
PURPOSE
What they should understand or do afterward.
WHAT I KNOW THAT THE INTERNET DOES NOT
Specific numbers, examples, situations I have seen,
opinions I hold and why.
CONSTRAINTS
Length. Format. Words never to use. Claims I cannot make.
GOOD LOOKS LIKE
[paste 200 words of your own writing]
STRUCTURE
Section 1: does X
Section 2: does Y
Section 3: does Z
Fill it once for a content type and reuse the top and bottom sections forever. Only the middle changes per piece.
The same request, twice
Take a bookkeeper who wants a page about catch-up bookkeeping.
The lazy version: "Write a blog post about catch-up bookkeeping services for small businesses."
What comes back is competent and useless. It explains what catch-up bookkeeping is, lists benefits any firm could claim, and ends by encouraging the reader to reach out to a qualified professional. Nothing in it is wrong. Nothing in it is theirs.
The briefed version:
AUDIENCE
Small business owners in Pittsburgh who are six to eighteen
months behind on their books. They are embarrassed about it.
Most have been avoiding dealing with it for a while.
PURPOSE
Reduce the embarrassment enough that they call.
WHAT I KNOW THAT THE INTERNET DOES NOT
- Most clients are 9 to 14 months behind, not 2 or 3
- Nobody has ever been judged for it in this office
- Two years of catch-up typically takes 3 to 4 weeks
- The real trigger is usually a loan application or an
IRS letter, not a New Year's resolution
- Clients consistently say the dread was worse than the work
CONSTRAINTS
900 words. No jargon. Never say "solutions" or "leverage."
Cannot promise penalty removal.
GOOD LOOKS LIKE
[200 words from their About page]
STRUCTURE
1. Name the feeling without being patronising
2. What actually happens, step by step, with timings
3. What it costs and what it depends on
4. What to gather before calling
That brief takes ten minutes to write and produces something the competitor's page cannot match, because the competitor's page does not contain any of those five facts.
Notice that none of the specifics required research. They came from doing the work and paying attention.
Where people get the brief wrong
Describing the topic instead of the reader. "Write about our automation service" is a topic. "Write for an owner who keeps losing leads because nobody replied for four days" is a brief.
Adjectives instead of examples. "Professional but approachable" means nothing consistent. Pasting 200 words of your own writing does more than any list of adjectives, because it demonstrates rather than describes.
Skipping the banned list. Two minutes writing down the words that make you wince saves rewriting every piece of output. This belongs in your brand standard so you write it once.
Not saying what you cannot claim. The model will produce plausible, confident, unsupportable statements if nothing stops it. In a regulated field this is the difference between a useful tool and a liability.
How to iterate without starting over
When output is not right, resist rewriting the request from scratch. Diagnose instead.
- Too generic? The "what only you know" section was thin. Add three more specifics.
- Wrong tone? Your example was not representative. Paste a better one.
- Wrong structure? Be explicit about what each section does, not just what it is called.
- Confidently wrong facts? You asked it to supply information you should have supplied. Never let it invent numbers.
That last one matters more than the others. Any figure in your content should have come from you or from a source you checked. A model will produce a plausible statistic with total confidence, and it will be your name on the page.
The part that makes this worth it
A good brief is reusable. Write one for blog posts, one for service pages, one for emails, and the marginal cost of the next piece collapses.
That is the actual productivity gain, and it is not the one people expect. The gain is not in the writing. It is in never again deciding how to say it, because the brief pre-decided that.
Storing those briefs so they get used is a separate small job covered here, and this covers the editing pass that catches what the brief missed.
Eric Howard
Founder of SteelAIQ and a chief marketing officer with more than twenty years running marketing organizations across enterprise software, manufacturing, and consumer brands. Based in Pittsburgh.
Related reading
AI Slop Is Real. Here's How Not to Publish It.
