1. Write your voice down before you brief the tool

You cannot keep a voice consistent if it only exists in your head. Before I let an AI draft anything for a brand, I write a one-page voice note. Not a forty-page brand bible — one page, with four things on it.

First, three adjectives that describe how the brand sounds: for one client it is “plain, warm, direct”; for another, “dry, confident, a little cheeky.” Second, three things the brand never says — the corporate phrases it refuses, like “leverage synergies” or “in today’s fast-paced world.” Third, two sample sentences written by a human, in the brand’s real voice. Fourth, the reader: who they are and what they need from you.

That page becomes the first paragraph of every brief. I paste it in, word for word, every time. It feels repetitive. That is the point. Consistency is repetition, done on purpose.

2. Brief with examples, not adjectives

Telling an AI to “sound professional but friendly” buys you almost nothing, because those words mean different things to different people — and to different models. Showing it works far better.

I give the tool two short samples of real brand writing — a paragraph from a post that landed well, and one from an email customers actually replied to — and I say: “Match this tone, this sentence length, this level of formality.” Then I give it the facts for the new piece, and nothing else. No vague mission statements, no “write something inspiring.”

Here is the difference in practice. A brief that says “write a friendly product update” returns a paragraph full of exclamation marks and the word “exciting.” A brief that includes a real past update — short sentences, one concrete detail, no exclamation marks at all — returns something a reader could mistake for the same author. The sample does the work the adjective only promises.

3. One voice check, on every draft, before it ships

Every AI draft I publish goes through the same three-question check, and it takes about four minutes.

First, could a reader tell this came from us without seeing the logo? If the answer is no, the draft is too generic, and I rewrite the opening and closing in my own words — those are the places voice lives or dies.

Second, does it contain any phrase from our never-say list? AI tools love the very phrases brands ban: “delve,” “in conclusion,” “it’s important to note.” I search for them and cut them. It is a short search and it catches most of the drift.

Third, does it sound like the same person who wrote last week’s post? I read the last paragraph of the previous post and the first paragraph of the new one, back to back. If they sound like strangers, something went wrong in the brief, and I fix the brief — not just the draft — so the next piece does not repeat the mistake.

4. Keep one human editor for every channel

The fastest way to lose a consistent voice is to let five people brief five tools in five different ways. One person writes the emails, another briefs the blog, a third hands the tool to an intern for social posts, and within a month the brand has three accents.

For a small brand, the answer is simple: one person owns the voice. They write the voice note, they approve the briefs, and they run the three-question check on everything before it publishes. That does not mean one person writes everything — it means one person guards the standard. When a new writer or freelancer joins, they get the voice note first, and their first three drafts get the full check, not a skim.

I have seen brands double their output with AI and lose their voice in the same month, and the cause was never the tool. It was that nobody owned the standard anymore. The tool just made the drift faster.

5. Update the voice note when the brand changes

A voice note is not a monument. Brands change: they get bolder, they drop a product line, they start talking to a different audience. When that happens, update the page — and, just as important, retire the old samples.

I review the voice note every few months and ask two questions: does this still sound like us, and do the samples still represent our best work? If a sample feels dated, I swap it for a recent piece that does. Stale samples are a quiet trap: the AI keeps matching a voice you have already outgrown, and your readers wonder why the brand sounds stuck in last year.

One practical habit helps here. Whenever a post gets an unusually warm response — a reply, a share, a comment that says “this sounds exactly like you” — I save that paragraph as a candidate sample. The voice note stays alive because real reader reactions keep feeding it.

The honest truth

Brand voice is not a setting you switch on in an AI tool. It is a document, a brief, and a check — three small habits, repeated every time. Write the voice down, brief with real examples instead of adjectives, check every draft against the standard, keep one person in charge of it, and update the note when the brand moves.

Do that, and AI becomes what it should be for a brand: a fast first draft in a voice your readers already trust. Skip it, and you will publish more than ever — and sound less like yourself with every post. Readers forgive a slow week. They do not forgive a brand that stops sounding like itself.

Frequently Asked Questions