Writing in a voice that is not yours, on purpose
Almost everyone who executes a brand voice did not invent it. Employees, ghostwriters and agencies all write as someone else, and voice guides are written as though that never happens. Here is the mechanics of doing it well.
Voice guides are written for the founder. They assume the person writing is the person the voice belongs to, and that the job is self-expression with some consistency applied.
Almost nobody is in that position. The social media manager did not choose the voice. The ghostwriter has never met most of the audience. The agency writer is holding four voices at once and switching between them on a Tuesday afternoon. Executing a voice you did not invent is the normal case, and the guides are written for the exception.
So: the mechanics. What actually makes it possible to write convincingly as somebody else.
The refusal list does almost all of the work
Here is the thing that surprises people the first time they try this seriously: the positive half of a voice guide is nearly useless to an executor, and the negative half is nearly sufficient.
"Warm but authoritative" gives a writer nothing to act on. Two competent people will produce completely different paragraphs from it and both will defend theirs.
"Never urgency that the market did not actually create" — a real line from one of those cards — is different in kind. It is checkable. A writer can look at a draft and determine whether they broke it, and so can a reviewer, and they will agree.
That asymmetry is why our voice examples print a refusal column beside every card and will not publish one without it. Look at what the refusals do:
- Never "secrets", "hacks", or anything the reader could not repeat to their accountant. — an adviser's voice, made operable.
- Never "handcrafted with love", which is the phrase that signals nobody was in the room. — a maker's voice, with the exact failure named.
- Never contrarian for its own sake. If you cannot show the receipt, do not run the post. — a heretic's voice, with the thing that turns it into noise fenced off.
Each one is a decision a writer can be wrong about. That is the entire test of whether a line belongs in a voice guide.
If you are executing a voice, ask for the refusals first. If nobody can produce any, the voice has not been decided and you are being asked to invent one while being held to it — which is the actual source of most "that does not sound like us" feedback.
Voice is a small set of habits, not a personality
The second thing that makes this tractable: a voice is much smaller than it feels. Most of it is four or five mechanical habits that you can hold in working memory.
- Sentence length and its variation. Does this voice run long and then stop short, or hold a steady medium?
- Who it addresses. "You" directly, "we" collectively, or nobody in particular. This is the single most detectable inconsistency and the one most often broken by accident.
- Where the point sits. First sentence, or arrived at?
- What it does with numbers. Leads with them, uses them as support, or avoids them.
- What it does with humour. Never, dry and occasional, or as the register.
Write those five down for the voice you are executing. That page will do more for you than the twenty-page brand book, and you can hold it beside the draft.
Read three real pieces, then write the fourth
Descriptions of a voice are lossy. The compression is severe and everything distinctive is in what got lost.
So before writing, read three real pieces in the voice — not the guide, the actual published work — and write immediately afterwards. You are pattern-matching, which is a thing humans are extremely good at and extremely bad at describing.
This is also the honest reason a captured sample beats a described one, and it is why the voice cards carry a caption written in the voice rather than only a description of it. The description tells you what to aim at. The sample tells you what hitting it looks like.
The failure mode is not sounding wrong. It is sounding like nobody.
When people write in a borrowed voice and it goes badly, they usually do not produce something that sounds like themselves. They produce something that sounds like no one — the safe midpoint of commercial English, hedged, competent, unattributable.
That happens because the writer is uncertain, and uncertainty averages. Every distinctive choice carries a risk of being wrong for this brand, so the risk-minimising move is to make no distinctive choices at all.
The fix is counterintuitive: be more specific, not less. A wrong distinctive sentence gets corrected in review and teaches you the boundary. A hedged neutral one passes review and teaches nobody anything, and forty of those is how an archive stops having an author. That accumulation is brand drift, and this is one of its two main sources — the other being genuine changes of mind, which are fine.
Consistency across writers is a checklist problem
If more than one person writes for the brand, taste will not hold it together. Taste is individual by definition, and each person's is applied to their own piece in isolation, which is exactly the blind spot drift lives in.
What holds it together is a short shared checklist, run per piece:
- Does it break any refusal?
- Is the address consistent — "you", "we", neither?
- Is the sentence rhythm in range?
- Would a regular reader attribute this to us with the logo removed?
Four questions, thirty seconds. Question 4 is the real one and the other three are how you answer it without arguing about taste. The same logic sits behind the on-brand content checklist: a checklist beats judgement here not because judgement is bad, but because judgement evaluates one piece and drift only exists across many.
When the voice is wrong
Sometimes you are executing a voice that does not work. The refusals contradict each other, or the voice does not fit the audience it is aimed at, and you can tell because the writing fights you every time.
Say so, and say it specifically: name the refusal that made the last three posts worse, and bring the drafts. "The voice is not working" is a taste argument nobody wins. "This refusal forced me into a worse opening on these three posts, here they are" is a design bug report, and it usually gets fixed.
If you are the one who defines the voice rather than executing it, the reciprocal obligation is to make it decidable: six branding words used to avoid a decision covers the terms most often used to avoid that, and the twelve voice cards are worked examples of what a decided one looks like.
Two adjacent problems this post does not solve. Executing a voice for a whole PROFESSION rather than a company has its own structural catch, and personal branding by profession takes fifteen jobs one at a time. And when the piece is an announcement rather than a regular post, the voice is not the hard part: we measured thirty announcement drafts and found the difference is not tone at all.