how to write in your brand voice

your brand voice isn't a mood board. it's a set of measurable writing habits. here's how to find them, document them, and keep ai from smoothing them into nothing.

the short answer

to write in your voice, you audit your 10 best pieces of writing. find the patterns. document four signals: sentence-length variation, vocabulary specificity, transition fingerprint, and structural habits. build a guide with examples, not adjectives. then score every draft against that profile to stop voice drift before it starts.

what brand voice is, really

brand voice is the set of patterns that makes your writing sound like you. even without a logo. it’s not a feeling, like “bold” or “authentic.” it's the math of your sentences. the kinds of nouns you use. the way you connect ideas. it’s why a paul graham essay sounds like him after a single paragraph.

your voice has three layers. most guides only see the first one. that's why they don't work.

  • the surface (the what): word choice, punctuation, jargon. easy to see. easy for ai to copy. think of ben settle signing off his emails with "your pal."
  • the structure (the how): sentence rhythm, paragraph shape, transitions, the flow of an argument. this is where your real voice is. the jagged pace of short and long sentences, not the words themselves.
  • the cognition (the why): your map of the world. how you get from a premise to a conclusion. this is the hardest part to copy. it's the last thing to die.

what voice guides get wrong

most brand voice guides are useless. they describe a personality, not a process. a notion page that says "we are witty and insightful" and "we are not dry and generic" gives a writer nothing to act on. how do you 'write wittier'? the guide has no answer. it's a mood board. you need a manual.

this leads to a predictable failure. the writer, told to be "witty," has no instructions. so they either write in their own voice, which creates chaos. or they ask an ai to "make this witty." the ai doesn't know your structural dna, so it spits out generic quippy text that sounds like everyone and no one.

what no one says about voice guides guides based on adjectives are worse than nothing. they pretend there's alignment but provide no way to get there. you can't act on "be more insightful." you can act on "every claim needs a data point or a named example."

step 1: audit what works

you don't invent a brand voice. you find the one you already have. the one that already got you readers and customers. that's the one that works. your only job is to document it.

gather your ten most resonant pieces of writing. resonance isn't likes or views. it's a specific human reaction.

  • an email from someone saying "that line hit me."
  • a comment that quotes you back to yourself.
  • emails with unusually high reply rates.
  • posts people share with their own thoughts attached.

ignore the duds. ignore the polite, generic praise. you're looking for the pieces where the signal was clear. this is your "resonance set." it's the only raw material that matters. everything else is noise.

step 2: find your patterns

with your resonance set, you're looking for four measurable signals. you're not reading for meaning. you're reading for structure. like an architect surveying a building, not a tourist reading the plaques. open a spreadsheet.

signal 1: sentence-length variation

paste your text into something like the hemingway app. look at the sentence lengths. are they all the same? or do you have a jagged rhythm of short sentences and long ones? a real voice has a high standard deviation here. a flat, corporate, ai voice has a low one.

signal 2: vocabulary specificity

this is the ratio of concrete nouns ("a porsche 911," "a google sheet") to abstract ones ("solutions," "synergy," "growth"). count them. a high ratio means you ground your ideas in real things. your writing feels credible. a low ratio feels vague. corporate.

signal 3: your transition fingerprint

how do you get from one idea to the next? list all your transition phrases. you'll find you only use a few, over and over. that's your transition fingerprint.

in the writing we analyze, people like justin welsh have a transition fingerprint of just 5-7 phrases they use 80% of the time. welsh uses "here's the thing" or "let's break that down" constantly. that repetition is a huge part of a recognizable voice. — hold your voice analysis, may 2026

signal 4: structural habits

document your repeating moves. how do you open? with a question? a data point? a story? how do you close? a summary, a call to action, a final question? how do you make an argument? problem-agitate-solve? numbered lists? these are your blueprints.

step 3: build a useful guide

now turn those patterns into a guide. a short doc in notion or google drive. it should be full of mechanical rules and examples. not aspirations.

a useful guide has three parts:

  1. the four signals: your exact numbers for sentence length, specificity, and your transition fingerprint.
  2. structural rules: your habits, turned into simple commands.
  3. before/after table: the most important part. show the rules in action.

here is what a practical rule looks like:

pattern the rule bad (before) good (after)
lead with the object start sentences with the benefit, not the company action. we've just shipped a new integration for our customers. your notion workflow just got faster. we shipped the new integration.
high specificity replace abstract nouns with concrete nouns. our platform offers robust solutions for enhanced productivity. our platform's zapier integration lets you log tasks in 1 click.
jagged rhythm follow a long sentence with a very short one. always. our team worked hard to analyze user data and found some interesting conclusions about their behavior. after analyzing three months of user data, we found a pattern. it was obvious.

how ai flattens voice

ai is the fastest way to kill a brand voice. the tools are built to produce average, predictable text. they sand down the very imperfections that make a human voice recognizable—the jagged sentence rhythm, the weird transitions, the specific vocabulary.

this isn't a theory. we've measured it.

in our analysis of 200+ voice profiles, writers using ai for drafts see a 60-70% drop in sentence-length variation within three posts. their standard deviation for sentence length—a key signal of a living voice—collapses from 12+ down to a robotic 4-6. — hold your voice voice profile dataset, 2026

the pattern is always the same. a writer uses an ai to draft something for speed. they edit it. it feels fine. they do it again. by the third draft, they've stopped writing in their voice and started editing the ai's. their signature opening is gone, replaced by "in today's fast-paced world..." their punchy sentences are gone, averaged into mush. readers don't know why, but the writing suddenly feels generic. soulless. this is voice drift. it happens in weeks, not years.

using ai without drift

you can use ai without losing your voice. but you need a specific workflow. the rule is simple: use ai for anything except the first draft. that first act of writing is where you embed your own patterns. outsource that, and the drift begins.

a safe way to use ai:

  1. ai for research. use it to brainstorm topics, generate questions, find data.
  2. ai for outlining. ask it for a structure. then edit it heavily until it follows your logic.
  3. human for the first draft. write the whole thing yourself. no ai. make it messy. make it fast. but make it yours. this is the only step that matters.
  4. ai for editing. once the draft is done, use ai to tighten a paragraph or suggest headlines. you're editing your voice, not its.
  5. score for drift. before you publish, run the final draft through a tool like hold your voice. score it against your profile. get a number that proves you still sound like you.

step 4: measure every draft

to protect your voice, especially on a team, a guide isn't enough. you need to measure it. you wouldn't run sales without a crm or engineering without automated tests. writing is the same. every piece of content needs to be scored against your voice profile before it's published.

this creates a real feedback loop. a writer submits a draft. instead of "this doesn't feel right," they get a score. "your sentence-length variation is a 5. we need a 12. break up the paragraphs. your specificity ratio is 0.3. our target is 0.6. replace 'solutions' and 'initiatives' with actual product names."

think like an engineering team the marketing teams we see with perfect voice consistency treat it like a code check. every post, email, and update is scored against their voice profile. if it fails the check, it doesn't ship. that's it.

this turns voice from an art into an engineering problem. it's the only way to keep a sharp voice when multiple people are writing, or when ai is in the mix. if you don't measure it, it drifts.

check your voice for drift

hold your voice builds a profile from your best writing, then scores every new draft against it. see your numbers in seconds. catch drift before your readers do.

start a free trial →

faq

what is brand voice in simple terms?

it's the set of measurable patterns—sentence rhythm, vocabulary—that makes writing sound like it came from you. it's not a list of personality traits.

is brand voice the same as brand tone?

no. voice is the stable structure. tone is the emotion you apply for a situation. voice is your architecture; tone is the paint job.

can ai writing tools really damage brand voice?

yes. our data shows using ai for first drafts cuts sentence-length variation by 60-70% in just three articles. ai defaults to average, and average erases you.

how do i find my brand voice if i'm just starting out?

you already have one. find writing you've done that got a real response—emails, posts, anything. analyze it for the four signals. your voice is in there. you just need to find it.

what are the most important things to include in a voice guide?

numbers, not adjectives. include targets for sentence length and specificity. list your top 5 transition phrases. and have a big table of before-and-after examples for your main rules.

how do you scale brand voice across a large team?

a guide helps. measurement scales it. create a central voice profile and score every piece of content against it before it goes out. it replaces subjective feedback with a number. it makes consistency an engineering problem, not a people problem.

shashank
ai
shashank

founder of hold your voice. writes about voice, ai patterns, and how to sound like yourself.

co-written with ai as sidekick. shashank built the four-signal framework. the ai helped structure the guide and stress-test the arguments.