most brand voice guides die in a notion doc. written once, shared once, and then quietly ignored while the actual writing drifts toward whatever pattern feels easiest that day. the problem is not that teams do not care about voice. it is that most guides describe voice at a level of abstraction that is useless at the sentence level, which is where voice either exists or it does not.
what does a brand voice guide actually need to contain?
four things: a voice description grounded in real examples, a short list of attributes with before-and-after sentences, a do-not-say list, and format-level tone notes. everything else is decoration. the guides that run to 30 pages with brand pyramid diagrams and personality wheels are monuments to process. writers do not consult them mid-draft because they cannot. you need something a person can scan in two minutes and immediately use.
start with what you already have. pull the five or ten pieces of writing that best represent how you want to sound. not the most polished or the best-performing by traffic, but the ones that feel most like you. blog posts, emails, product copy, social posts — whatever format you actually produce. these samples are where everything else comes from.
how do you extract voice attributes from existing writing?
read your samples out loud and note what they share at the sentence level. the patterns you are looking for are structural: sentence length, punctuation habits, how often you use questions, whether you state opinions directly or qualify them, whether your examples are specific or vague. you can also use a tool like the brand voice analyzer to measure these dimensions across a set of samples rather than relying on feel alone.
once you see the patterns, name them as attributes. three to five is enough. more than five and writers cannot hold them in working memory. each attribute needs two things to be usable: a one-sentence description and a concrete example. "direct" as an attribute means nothing on its own. "direct: we say 'this will not work for you if' rather than 'this solution may not be the ideal fit for all use cases'" is something a writer can act on immediately.
the examples should come from your actual writing, not from sentences you invented to illustrate the point. if you invented them, they are aspirational. if they came from real work, they are descriptive. aspirational voice guides describe a brand nobody has actually been yet. descriptive ones codify what is already working.
what should go on the do-not-say list?
the do-not-say list is often the most useful part of the guide because it is the most specific. it should include words and phrases that contradict your voice or have been drained of meaning through overuse. words like "leverage," "synergy," "holistic," "best-in-class," and "game-changing" belong here for most brands. so do filler phrases like "it goes without saying," "in today's world," and "at the end of the day."
this list matters more now than it did five years ago. when writers use ai tools to draft or edit content, the output defaults toward the most statistically common patterns in training data — which means it defaults toward exactly the overused phrases your list should block. a do-not-say list gives writers something concrete to check ai-generated drafts against. a filter, not just a style preference.
add a second column: the replacement. not just "don't say leverage" but "say use instead." not just "avoid best-in-class" but "say what specifically makes it better." the replacement column turns the list from a prohibition into a writing instruction.
how is tone different from voice, and does your guide need to address it?
voice is fixed. tone shifts. your voice is the consistent personality behind everything you publish — the opinions you are willing to state, the sentence structures you favour, the things you never say. tone is how that personality adjusts for context. a support email and a product launch announcement should both sound like you, but one is warmer and the other is more direct. your guide needs to acknowledge this or writers will either flatten tone entirely or push it so far in one direction that the voice disappears.
a simple format-level section is enough. list the content formats you produce most often, and for each one note one or two ways tone adjusts without the core voice changing. "blog posts: we are willing to take strong positions and use contractions freely. support emails: we lead with empathy and keep sentences short. product announcements: we are specific about what changed and why, no hype." that is all you need. if you are seeing voice drift across formats, this section is usually where the fix starts.
how long should the guide be, and who should write it?
two to four pages. if you cannot capture your brand voice in four pages, the problem is not length — it is that you do not have a clear enough picture of what your voice actually is. a 30-page guide signals internal disagreement about voice that nobody wanted to resolve, so they added more sections instead.
the guide should be written by whoever does the most writing for the brand, not by a committee. committees produce documents that describe the intersection of multiple people's preferences, which is not a voice — it is a negotiation. get input from others, but give one person the final call on every sentence in the guide. that person's job is to be descriptive about what works, not diplomatic about what everyone wants.
how do you keep a brand voice guide from going stale?
review it when your writing changes, not on a calendar schedule. the trigger for a review is noticing that your best recent work no longer matches the guide — that the guide describes an older version of how you write. this is a good problem to have. it means your writing is developing. update the samples first, then update the attributes to match.
the bigger risk in most teams right now is not that the guide goes stale but that it gets ignored because ai tools make it easy to produce volume without checking consistency. if multiple people are using ai-assisted drafting, voice drift can accumulate faster than any editor can catch it manually. running drafts through a voice consistency check before they publish is the only way to catch the drift at scale. the guide tells writers what the voice is. the check tells you whether the draft actually has it.
a brand voice guide is not a document you write and file. it is a working definition of what makes your writing yours. that definition only has value if the writing you ship matches it. building the guide is step one. the harder step is building a process that keeps the gap between the guide and the published work from widening every time a deadline hits or an ai tool gets into the drafting flow. read more about what brand voice actually means before you start writing the guide, and you will write a better one.