ع
Start Topics Teams Reference What's new Saved
playbook

Build a brand-voice system Claude can actually use

Reverse-engineer your voice from the writing you're proud of into a reusable doc — then use it to keep every future draft, from any teammate, on-brand.

medium ~1 hour to build, reused forever
when to reach for this

"Write it in our brand voice" is meaningless to Claude — and honestly to most of your team — unless the voice is written down. The fix isn't to write an adjective list ("friendly, bold, human"); it's to extract concrete, testable rules from copy you already love, into a brand-voice.md you paste into every content task. Build it once and every other Marketing playbook gets sharper, because "our voice" finally points at something.

gather this first
  • 3–6 pieces you're proud of — posts, emails, landing pages. Paste them in or save as files.
  • 1–2 pieces that felt OFF-brand (a draft a vendor sent, an old post that makes you wince). Counter-examples teach as much as examples.
  • Any existing style notes, however rough.
the workflow

  1. Extract the rules from your best writing

    Save your sample pieces into a folder, open it in Claude Desktop, and ask in the chat — no terminal needed. Have Claude infer the patterns rather than you trying to articulate them. It's much easier to recognize your voice than to define it from scratch.

    you ask
    Here are 5 pieces of our writing I'm proud of. Reverse-engineer our voice into rules: sentence length and rhythm, vocabulary we use and avoid, how we open and close, our stance on jargon/emoji/exclamation points, and the feeling a reader should walk away with. Be specific and testable, not vague adjectives.

    what you get back A concrete rule set — "short declarative sentences, one idea each; no exclamation points; we explain before we name; we never say 'leverage' or 'seamless'" — grounded in your actual copy.

  2. Sharpen it with counter-examples

    Feed in the off-brand pieces and ask what makes them wrong. Negative space defines the voice as much as the positive rules.

    you ask
    Here are 2 pieces that feel OFF-brand for us. For each, name exactly what breaks our voice, and add 'we don't…' rules to the voice doc so a future draft avoids it.

    what you get back A do/don't list with specific failure modes ("we don't open with a rhetorical question; we don't stack three adjectives") — the guardrails that catch drift.

  3. Assemble the reusable doc

    Pull it all into one clean artifact, with examples baked in. The examples are the most important part — they're what Claude actually pattern-matches on.

    you ask
    Assemble everything into a clean brand-voice.md: a one-paragraph summary, a numbered rule list, a do/don't table, and 3 short before/after rewrites showing a flat line turned into our voice. Keep it under one page.

    what you get back A single, paste-ready brand-voice.md — summary, rules, do/don'ts, and worked before/afters.

    Save this file. It becomes the first thing you paste into every content task from here on.

  4. Battle-test it on a cold draft

    Prove the doc works by having Claude apply it to something new, then to something deliberately bad, so you can see it catch and correct.

    you ask
    Using only brand-voice.md, rewrite this flat product paragraph in our voice — then show me a version that VIOLATES three of our rules and label which rules it breaks. I want to see the doc working in both directions.

    what you get back An on-voice rewrite plus an annotated off-voice one — proof the doc is specific enough to both guide and catch.

make it your own
  • Per-channel voices: extend the doc with a short LinkedIn vs X vs email section if your tone shifts by channel.
  • Arabic & bilingual voice: if you publish in Arabic, build a parallel brand-voice-ar.md from Arabic copy you're proud of — don't translate the English rules. Arabic carries its own register, rhythm, and formality; a line that lands in English can read stiff or off-key in Arabic. Two voice docs, one brand.
  • Onboard humans too: the same doc is the best onboarding for new writers and freelancers — it's not just for Claude.
  • Make it a skill (Power Track): advanced teams can turn brand-voice.md into a reusable skill or /onbrand command so it's always loaded (see the Capabilities tab).
watch out for
  • Adjectives are not rules. "Bold and human" gives Claude nothing; a before/after rewrite gives it everything. Bias the doc toward examples.
  • One canonical doc, one owner. Across a team, a voice doc that everyone edits quietly drifts. Keep a single source of truth, version it, and treat changes as a reviewed update — not a free-for-all every freelancer can "improve."
  • Refresh it when your voice genuinely evolves — but don't let it churn every week, or it stops being a stable reference.
  • A voice doc makes drafts sound right; it doesn't make claims be right. Still fact-check, and route anything regulated or legally loaded past the people who own that sign-off.

you'll end up with A one-page `brand-voice.md` that makes "our voice" a concrete, testable thing — sharpening every other content task and onboarding humans and Claude alike.

Questions people ask

What if I can't articulate my brand voice in the first place?
That's the point of the system — you don't have to. You paste in 3–6 pieces you're proud of and Claude reverse-engineers the rules. Recognizing your voice in real writing is far easier than defining it from a blank page.
How is a `brand-voice.md` better than a list of adjectives like \"friendly, bold, human\"?
Adjectives give Claude nothing testable — "bold" can't catch a bad sentence. A `brand-voice.md` built from concrete rules and before/after rewrites can: it tells Claude exactly which words to avoid, how to open and close, and what an off-brand line looks like.
Will the voice doc make my drafts factually correct?
No. A voice doc makes drafts sound right; it does not make claims be right. You still fact-check every statistic, name, and promise — the doc governs tone, not truth.
Can the same doc onboard new writers and freelancers, not just Claude?
Yes, and it's one of the best reasons to build it. The same concrete rules and examples that keep Claude on-brand are the fastest way to get a new human writer or vendor sounding like you.