Est.

Brand Drift Diagnostics · Template

Brand voice guidelines template, with examples

A free, copyable template for writing down how your brand sounds: voice traits turned into rules and yes/no tests, a tone map by context, word lists, before-and-after examples, and a block of instructions to give the AI tools that write for you.

Why “friendly, confident, human” isn’t a voice guide

Most voice guidelines are a row of adjectives. A writer who has been on the team for years reads “confident” and knows what it means here. A new hire, a freelancer or a model reads the same word and fills it in from everywhere else, which is how so much AI copy ends up sounding like everyone else’s.

A voice is only written down once each trait has a limit, a rule you can follow and a test two reviewers would answer the same way. That is also what lets you enforce it across AI tools rather than hope for it. This template is built around that.

What the template covers

  1. Our voice in one sentence: who you sound like, talking to whom, about what: the line every other rule should agree with.
  2. Voice traits: three or four traits, each with the excess it must not tip into, a do, a don’t and a yes/no test.
  3. How we write sentences: person, sentence length, contractions, hedging, openings and punctuation: where voice actually lives.
  4. Tone map: what turns up and what turns down in each context, with a length limit, so the voice holds while the tone moves.
  5. Words: exact names, we-say/we-don’t-say pairs, and the words you never use.
  6. Examples: before-and-after pairs, one per context, each naming the rule it applies.
  7. Instructions for AI tools: the same rules written as instructions a model can follow and check its draft against.
  8. Review checklist: the checks a piece passes before it ships, each pointing back to a section.
  9. Changes: a version, an owner and a changelog, so everyone knows which voice is current.

The template

Markdown, so it pastes cleanly into a doc, a wiki, a repo or an AI tool. Replace everything in angle brackets.

# Brand voice guidelines: <Brand name>

Version: <1.0>   Owner: <name, role>   Last changed: <date>
Applies to: <every channel / list the products, regions and teams>
Part of: <link to your full brand guidelines, if you have them>

## 1. Our voice in one sentence
We sound like <who: e.g. "a senior colleague who has done this before">
talking to <reader> about <subject>: <two or three words for how>.

## 2. Voice traits
Three or four traits. Each one has a limit (what it must not become),
a rule anyone can follow, and a test a reviewer can answer yes or no.

### <Trait 1>, not <its excess>
- Means: <one sentence in plain words>
- Do: <concrete instruction>
- Don't: <concrete ban>
- Test: <yes/no question a reviewer can answer from the text alone>

### <Trait 2>, not <its excess>
- Means: <...>
- Do: <...>
- Don't: <...>
- Test: <...>

### <Trait 3>, not <its excess>
- Means: <...>
- Do: <...>
- Don't: <...>
- Test: <...>

## 3. How we write sentences
- Person: <we / I> for us, <you> for the reader; never <the user / customers>
  when speaking to them
- Sentence length: <average under N words; none over N>
- Contractions: <yes, always / only in <channels> / never>
- Hedging: <words we cut, e.g. "might", "arguably", "in some cases",
  unless the uncertainty is real and stated>
- Openings: <lead with the outcome / the answer / the reader's problem>
- Punctuation: <exclamation marks: never / max one per piece>;
  <emoji: never / only in <channels>>
- Jargon: <terms we always explain on first use>

## 4. Tone map (the voice stays; the tone moves)
| Context          | Turn up          | Turn down        | Length / format   |
|------------------|------------------|------------------|-------------------|
| <Product UI>     | <clarity>        | <personality>    | <max words>       |
| <Error / outage> | <ownership>      | <humour>         | <max sentences>   |
| <Support reply>  | <warmth>         | <selling>        | <max sentences>   |
| <Social>         | <...>            | <...>            | <...>             |
| <Sales email>    | <...>            | <...>            | <...>             |
| <Long-form post> | <...>            | <...>            | <...>             |

## 5. Words
- Our names, exactly as written: <Product Name>, <Feature name>
  (never <wrong forms>)
- We say / we don't say:
  <...>  not  <...>
  <...>  not  <...>
- Words we never use: <list, e.g. "seamless", "revolutionary", "leverage">
- Words that sound like us: <list>

## 6. Examples (before and after)
One pair per context in the tone map. Name the rule each fix applies.
- Context: <Product UI>
  Before: "<off-voice sentence>"
  After:  "<rewritten sentence>"
  Rule:   <trait or line in section 3 or 5>
- Context: <Support reply>
  Before: "<...>"
  After:  "<...>"
  Rule:   <...>

## 7. Instructions for AI tools
Paste this block, or point the tool at the current version of this file.
- Write as <Brand name>: <the sentence from section 1>.
- Follow every Do and Don't in section 2 and every rule in section 3.
- Use the tone row for <context> from section 4, within its length limit.
- Use only the names and words in section 5; never use the banned list.
- Do not state prices, dates, numbers, guarantees or customer names
  unless they are given to you in this task.
- Before answering, check the draft against the tests in section 2 and
  rewrite any sentence that fails.

## 8. Review checklist
- [ ] Every test in section 2 passes
- [ ] Sentences follow section 3 (person, length, hedging, openings)
- [ ] Tone and length match the context's row in section 4
- [ ] Names spelled and cased as in section 5; no banned words
- [ ] Two reviewers would give the same answer on each check

## 9. Changes
| Version | Date   | What changed     | Approved by |
|---------|--------|------------------|-------------|
| <1.0>   | <date> | <first version>  | <name>      |

Brand voice examples: one filled in

Here is section 2 completed for an invented invoicing app, so you can see the level of detail that makes a trait usable. The company and its copy are made up for illustration; copy the shape, not the words.

TraitDoDon’tTest
Direct, not curtPut the answer or the outcome in the first sentence.Open with “We’re excited to…”, “In today’s world…” or a question.Could a reader stop after the first sentence and know what happened?
Plain, not simplisticUse the word the reader would use; explain any term on first use.Use “leverage”, “seamless”, “robust” or internal feature codenames.Is every term one the reader would type into a search box?
Calm, not coldState problems and fixes in the same sentence; say sorry once, when it’s ours.Use exclamation marks, “unfortunately”, or more than one apology.Does the piece read the same whether the news is good or bad?

Before and after

The same three traits applied to real situations. Each fix names the rule it follows, which is what turns an example into training for the next writer or the next prompt.

Product UI (error)

  • Before: “Oops! Something went wrong. Please try again later.”
  • After: “We couldn’t send this invoice because the client’s email address is missing. Add it and send again.”
  • Rule: Direct: say what happened and what to do, in that order. Calm: no “Oops!”.

Support reply

  • Before: “Unfortunately, as per our policy, refunds cannot be processed after 30 days. We apologise for any inconvenience this may cause.”
  • After: “You’re outside the refund window for this one, so I can’t refund it. What I can do is cancel the renewal now so you’re not charged again.”
  • Rule: Calm: one fact, then the fix. Plain: no “as per”. Tone map: warmth up, policy language down.

Launch post

  • Before: “We’re thrilled to announce our revolutionary new feature that seamlessly transforms how you get paid!”
  • After: “You can now send a reminder automatically when an invoice is seven days late.”
  • Rule: Direct: the outcome first. Words: “revolutionary” and “seamlessly” are on the banned list.

How to write your brand voice guidelines

  1. Collect twenty pieces you’d publish again. Take them from different channels and writers. Your voice is already in there; the job is to describe it, not invent it.
  2. Collect ten you wouldn’t. Off-voice pieces are more useful than good ones: each one shows a rule nobody wrote down. AI drafts your team rewrote are a fast source.
  3. Name the traits from the gap between the two piles. Ask what the good pieces do that the bad ones don’t. Give each trait a limit, so “direct” doesn’t become rude and “playful” doesn’t become flippant.
  4. Write a test for every trait. If you can’t phrase a yes/no question a reviewer could answer from the text alone, the trait is still an adjective. Keep rewriting until you can.
  5. Move down to sentences. Person, length, contractions and hedging are where readers actually hear a voice, and where models drift first. Section 3 is often the most useful part of the whole document.
  6. Fill the tone map from your real contexts. List the places you write (product, errors, support, social, sales, long-form) and decide for each what turns up and what turns down.
  7. Test it on someone new. Give the guidelines to a new writer and to an AI tool, ask both for the same piece, and run the checklist. Where they miss, a rule is unclear. Fix it, bump the version and log the change.

Common mistakes

  • Traits that every brand could claim. “Friendly” and “professional” describe nearly everyone. The rules and tests are what make a trait yours.
  • A tone map with no length limits. Length is the easiest tone rule to check and the one models break most often.
  • Examples without the rule. A before-and-after with no rule named teaches one sentence. Naming the rule teaches every sentence like it.
  • Writing it once and letting it decay. A voice pasted into long prompts and workflows wears away over time. Version it, and check output against it on a schedule; see measuring brand consistency.

Where voice guidelines fit

Voice is one part of a brand guide. If you also need rules for claims, grammar and formatting, and visual identity, start from the brand guidelines template and use this page for its voice and tone sections. Either way, keep one versioned source that every writer and every AI tool reads from, rather than a copy in each prompt.

Questions

What are brand voice guidelines?

A short document that says how your brand sounds in writing, in a form people and tools can follow: a one-sentence description of the voice, three or four traits with rules and tests, sentence-level rules, how tone shifts by context, the words you use and avoid, and labelled examples.

What is the difference between brand voice and tone?

Voice is the personality that stays the same everywhere. Tone is how that voice adjusts to the moment: an error message and a launch post should both sound like you, but not the same. The tone map in section 4 is where you write those adjustments down.

How many voice traits should a brand have?

Three or four. Fewer rarely covers enough; more starts to contradict itself, and a reviewer can’t hold them all in mind. Each one needs a limit (“direct, not curt”), or a writer or a model will push it too far.

How do I get AI tools to write in our brand voice?

Give them rules they can check, not adjectives. Section 7 of the template turns the voice into instructions, including a final step to test the draft against each trait. Keep one current version that every tool reads, rather than a pasted copy in each prompt that drifts out of date.

Do brand voice guidelines replace a full brand guide?

No. They cover how you sound. A full brand guide also covers claims that need approval, grammar and formatting, and visual identity. If you need all of it, start from the brand guidelines template and use this page to write its voice and tone sections in more depth.