Brand voice guidelines need six working parts: a one-line description of the voice, a word list with real replacements, formality rules that shift by context, grammar and mechanics, example passages, and a named owner. Most brand guides give voice a paragraph and three adjectives instead, which describes how the brand feels without saying what to actually type. A new copywriter still has to guess, and so does an AI tool asked to draft a launch email on the brand's behalf.
This post goes through what each part is doing in a set of brand voice guidelines, why a single formality setting rarely covers a whole brand, and where voice fits inside the larger guide it belongs to.
Key Takeaways
- Brand voice guidelines need six parts: a one-line voice, a word list with replacements, formality by context, grammar and mechanics, example passages, and a named owner.
- Adjectives describe how a brand feels; a word list and a formality rule describe what to actually write, which is the part a stranger or an AI tool can check.
- Voice is one part of a brand's guide, next to logo, color, and typography, not a document on its own.
- Formality needs a separate line for each audience a brand actually writes to. One rating rarely covers customers, developers, and a careers page equally well.
- TrueKeep keeps brand voice guidelines in a design brief and connects it to Claude, ChatGPT, and Cursor through an MCP connection, so AI tools read the rules before they draft.
What makes brand voice guidelines useful
A set of brand voice guidelines is only as good as the test it passes: hand a sentence to someone who has never met the brand, and see if they can tell whether it belongs. "We sound friendly and professional" gives that person nothing to check against. A line like the one below gives them an answer in seconds, because there is something concrete on the page to compare the sentence to:
Never: unlock, elevate
Headlines: sentence case, eight words or fewer
The same test now applies to a second reader. When someone asks Claude or ChatGPT to write a support reply, the draft comes from whatever voice information the model has, which by default is none. A rule written clearly enough for a new hire to follow turns out to be exactly what a model needs before it writes anything.
The six parts, and what each one is doing
- The voice in one line. Not three adjectives. A short phrase that sets the posture, such as "confidence without swagger" or "direct, never cute."
- A word list. Words the brand reaches for and words it drops, each one paired with what to say instead, not just a warning label.
- Formality by context. How formal the brand is with customers, with developers, on a careers page, and in support replies, written separately for each.
- Grammar and mechanics. Sentence case or title case, the serial comma or not, how numbers are written, whether exclamation marks are ever allowed.
- Example passages. Two or three short passages that sound right and one that sounds wrong, each with a line explaining why.
- Ownership. The person who approves new words, retires old ones, and reviews the list on a schedule.
The middle three parts are the ones a stranger can act on without asking anyone a question. A short excerpt carries more weight than a paragraph describing "brand consistency":
Never: cutting-edge, robust, unlock
Use instead: name the feature and what it actually does
Headlines: sentence case, no exclamation marks
Numbers: spell out one to nine, numerals for 10 and up
A single formality setting rarely covers a whole brand
The line that works on a careers page ("tell us what you're proud of and why you want in") reads loose in a security questionnaire, and the line that works in a security questionnaire ("access is logged and reviewed on a schedule") reads cold on a careers page. The Nielsen Norman Group's research on tone of voice separates these two ideas directly: voice holds steady across a brand, while tone moves with whoever is reading at that moment and what they need from the page.
Brand voice guidelines that only state one formality rating have usually answered for customers and left developers, new hires, and support out of it. Writing a line for each, plain and direct with customers, precise and jargon-free with developers, warmer on a careers page, short and honest in support replies, costs one more paragraph and closes most of that gap.
Where voice sits inside a brand's larger guide
Voice does not stand alone. TrueKeep organizes a brand's rules across four parts of one portal, strategy, guide, assets, and components, and voice lives inside the guide alongside logo use, color, and typography. It also follows from the strategy above it: what a brand sounds like depends on who it serves and what it is trying to be, not the reverse. The brand guidelines template covers the full set of sections a guide needs; voice is usually the one written last and checked first.
That order has a reason. A brand manager asking "is this on-brand?" tends to check the words before the hex value, because a wrong headline is easier to spot than a wrong shade of blue. Voice guidelines that fit on one page, rather than a chapter a reader has to work through, are what keep that first check fast.
Writing voice guidelines an AI tool can actually read
A word list and a formality rule help a new hire write their first draft. The same two things are what a connected AI tool needs before it writes anything at all. TrueKeep calls this written-down version of the brand a design brief, and connects it to Claude, ChatGPT, or Cursor once through the AI connection guide, using the Model Context Protocol, an open standard for giving AI tools access to a source of information. From then on, the model reads the word list and the formality rules before it drafts, instead of working from the brand name alone.
For people writing inside Claude for most of the day, the brand kit skill brings the same brief into the conversation directly, so the voice rules are there without anyone stopping to look them up first.
A design brief only helps the drafts someone actually checks against it. TrueKeep runs a final pass over AI output before it ships, but a voice guide that no tool is connected to does as little for a model as a PDF nobody opens does for a new copywriter.
Keeping voice guidelines true after they're written
Brand voice guidelines are accurate on the day they're finished and start drifting the next day a product line launches or a word falls out of use. Nothing in the document says so until a reader notices a sentence that no longer fits.
Name one person in the ownership section, not a team, and give them a way to change the list directly instead of filing a request through someone else. Edit mode lets a curator retire a word the day it stops working, rather than waiting for a review that might be months out.
FAQ
What belongs in brand voice guidelines? Six parts: a one-line description of the voice, a word list with replacements, formality rules written separately for each audience, grammar and mechanics, example passages, and a named owner. Each part should give a reader something to check a sentence against, not just a feeling to match.
How is brand voice different from tone? Voice stays constant: the words a brand reaches for and the formality it defaults to. Tone moves with the situation. A security questionnaire reads differently from a careers page, while both stay inside the same voice.
Should brand voice guidelines be their own document, or part of a bigger one? Usually part of a bigger one. Voice follows from the same strategy that sets the logo, color, and typography rules, so most brands keep it as a section of the full guide rather than a separate file that can drift out of sync.
Does brand voice change after a rebrand? Often partly. A rebrand can retire specific words or shift formality without changing the one-line description of the voice itself. Checking the word list and example passages against the new brand is the part most teams forget to do.
How do I connect brand voice guidelines to Claude or ChatGPT? Put them somewhere the model can read before it drafts, rather than pasting them into each conversation. TrueKeep keeps the voice guide in a design brief, connected through an MCP connection, so the word list and formality rules are already there when someone asks Claude for a draft.
Who should own brand voice guidelines? One named person, usually whoever owns the full brand guide. A team-level owner means no one schedules the review, and the word list falls behind the brand it's supposed to describe.
The short version
Brand voice guidelines need six parts: a one-line voice, a word list with real replacements, formality written separately for each audience, grammar and mechanics, example passages, and a named owner. Skip the adjectives, since a word list and a formality rule are what a stranger, or an AI tool, can actually check a sentence against. Voice belongs inside the brand's larger guide, not off on its own, and it does the most good where both a person and a connected AI tool can find it in seconds.
To keep your brand voice guidelines in a portal your team and their AI tools both read, get started with TrueKeep.