Blog › Product

How to brief an AI writing assistant so it sounds like you

AI writing sounds generic when it has nothing specific to work from. What to put in a brief: audience, samples, rules and facts, with a template you can reuse.

Nov 14, 2026 · 5 min read

To make an AI writing assistant sound like you, give it four things: who you are writing for, examples of your own writing, clear rules about what to do and avoid, and the facts it needs. Of these, examples matter most. An assistant given three of your real posts will match your voice better than one given three paragraphs of instructions.

Generic output is nearly always a briefing problem, not a tool problem.

Why AI writing sounds generic

A language model, given nothing specific, produces the most typical version of whatever you asked for. Ask for "a LinkedIn post about leadership" and you get the average of every LinkedIn post about leadership. It is fluent, and it is nobody's.

The only cure is specific input. The model needs something to hold on to that is yours.

The four parts of a good brief

1. Audience

Say exactly who will read it.

• Weak: "professionals".
• Better: "founders of software companies with under twenty people, who are doing their own marketing".

Add what they already know and what they want. A model writes very differently for a beginner and an expert.

2. Samples

Paste three to five things you wrote and are proud of. This is the highest-value part of any brief.

Choose samples that:

• Sound the way you want to sound. Not your most formal piece, unless that is the goal.
• Are the same kind of thing you are asking for. Posts teach posts.
• Did well with the people you want to reach.

Put the strongest first. Tools often give most weight to the beginning.

No samples yet? Write one piece carefully by hand, or describe a voice you admire and why, then replace it with your own writing as soon as you have some.

3. Rules

Short, concrete, testable.

Do:

• Short sentences.
• One concrete example per post.
• Start with the point.
• Use real numbers where we have them.

Don't:

• Emojis.
• Hashtag walls.
• "Game-changing", "unlock", "in today's fast-paced world".
• Rhetorical questions as openers.
• Claims we cannot support.

A "don't" list removes more genericness than a "do" list adds, because generic writing is made of stock phrases.

4. Facts

A model does not know your prices, your product's limits or what happened last week. If the piece depends on facts, supply them.

• What the product does and does not do.
• Prices and plans.
• Policies: refunds, delivery, support hours.
• The specific story, number or decision this piece is about.

Without facts, the assistant fills gaps with plausible guesses. Those are the errors that embarrass you.

A reusable brief

AUDIENCE: [who, what they know, what they want]

GOAL: [what the reader should do or understand]

VOICE: [three words, each with a limit: direct but not blunt]

DO: [five rules]
DON'T: [five rules]

SAMPLES:
[three to five of your own pieces]

FACTS FOR THIS PIECE:
[the story, the numbers, the link]

FORMAT: [length, structure, platform]

Fill in everything above FACTS once and save it. Then each new piece needs only the last two parts.

Brief the piece, too

The standing brief covers how you sound. Each piece needs its own short instruction.

• Weak: "Write a post about our new feature."
• Better: "Write a LinkedIn post announcing reviewer roles. The angle: customers asked for a way to let an editor approve posts without admin access. Mention it took longer than expected because approval touches everything. End with where to find it. Under 150 words."

The better version gives an angle, a detail and a limit. That is the difference between a draft you can use and one you rewrite.

Tell it your stances

Voice is also what you believe. If you have positions, such as never naming competitors or always being upfront about pricing, state them. Otherwise the assistant will take the safe middle on everything, and safe middles are forgettable.

Edit like an editor

Even a good brief produces a draft, not a finished piece.

1. Check every fact. Names, numbers, dates, claims.
2. Cut the first sentence if it is throat-clearing.
3. Add one detail only you know.
4. Read it aloud. Change anything you would not say.
5. Feed the correction back. If you keep deleting the same phrase, add it to the don't list.

Step 5 is how the brief improves. Each edit is a rule waiting to be written.

What an assistant cannot do

• Have your experience. It can shape a story. It cannot supply one.
• Know what is true about your business unless you tell it.
• Judge timing and context.
• Take responsibility. What is published is yours.

See E-E-A-T: content that shows real experience.

How this maps to FinalPost AI

In FinalPost AI you do not paste a brief each time. The standing parts live in brand voice and apply to every draft.

• Audience and tone: choose a tone and describe who you want to reach.
• Samples: paste them, or use Learn from my accounts, which reads up to 25 recent posts from your connected channels, saves the three to five most representative as samples, and suggests do and avoid lists from patterns it actually found. Agents read roughly the first 2,000 characters of your samples, so put the strongest first.
• Rules: do use and don't use lists. You can add one by telling the assistant "stop using emojis"; after you confirm, it is saved for every future post.
• Stances: your positions with reasons, each marked normal or high risk.
• Facts: quick answers and documents about you and your work.

The per-piece part is the agent's goals, or your request to the assistant. And every draft waits for your approval, which is where editing happens.

One thing it will not do: use your Job Scout CV or application drafts as sources for public posts. Only what you add to Brand Studio is used.

To write the standing parts, start with how to find your brand voice.

Frequently asked questions

How many writing samples does an AI need?

Three to five good ones are enough to capture a voice. More helps little, and weak samples hurt. Put your best first.

Why does AI writing all sound the same?

Because without specific input a model produces the most typical version of what was asked. Samples, rules and facts give it something of yours to work from.

Is a longer prompt always better?

No. A clear brief with good samples and a few firm rules beats a long list of vague instructions. Length is useful only when it adds specifics.