Brand Voice Examples: Six Voices Written So People and AI Can Use Them

Most brand voice examples stop at three adjectives and a famous logo. These six show the whole voice: the traits, the same message written in it, the rules that produce it and the mistakes that mean it is slipping.

Updated

What a useful brand voice example shows

"Friendly, confident, human" describes an impression, not a voice. Hand those three words to a new copywriter and they will ask for samples. Hand them to a language model and it fills the gap with how companies in your category usually sound, which is how so much AI copy ends up interchangeable.

An example earns its place in a voice guide when it shows four things:

  • Traits as pairs. "Confident, not boastful" draws a line a writer can check a sentence against; "confident" on its own does not.
  • One message, rewritten. Seeing the same news in different voices makes the differences concrete in a way descriptions never do.
  • The rules behind it. Sentence length, contractions, words to use and avoid, when humour is off. These are what a model can actually follow.
  • What drift looks like. The specific ways the voice goes wrong, so an editor or an automated check knows what to flag.

Every example below follows that shape. For why adjectives fail a machine reader in particular, see What a Machine-Readable Tone Guide Contains.

Voice and tone are not the same thing

Voice is the character a brand keeps everywhere. Tone is how that character adjusts to the moment: a launch announcement, a billing error and a security notice should all sound like the same company, but not in the same register.

The examples here are voices. Each one notes where its tone should shift, because that is where most voices break, especially in AI tools that apply one setting to every situation. Separate Tone Variants Within a Single Brand Voice covers how to define those shifts as named variants.

Six brand voice examples

Each voice rewrites the same announcement from a fictional reporting tool, so you can compare them line for line. They are illustrations to borrow from, not descriptions of any real company's voice.

The plain version: “PDF export is now available for scheduled reports.”

  1. Example 1

    Plain-spoken expert

    Suits: B2B software, finance and tools for professionals who want the point first.

    We areWe are not
    Directcurt
    Knowledgeablelecturing
    Confidentboastful
    “Scheduled reports now export to PDF. Pick the format when you set the schedule, and the file lands in the same inbox it always did.”

    Rules that produce it

    • State the change in the first sentence.
    • Keep sentences short; split any sentence that carries two ideas.
    • Contractions are fine.
    • No intensifiers: "very", "incredibly", "game-changing".
    • Give one concrete next step.

    Drifting when: Opens with "We're excited to announce", stacks adjectives, or explains things the reader already knows.

  2. Example 2

    Warm and encouraging

    Suits: Consumer apps, education, wellness and community products.

    We areWe are not
    Friendlygushing
    Encouragingpatronising
    Personaloverfamiliar
    “Good news for your Monday: scheduled reports can now arrive as PDFs. Set it up once, and that's one less export to do by hand.”

    Rules that produce it

    • Speak to the reader as "you".
    • Name what the reader gains before how the feature works.
    • At most one exclamation mark per piece, and none in errors or billing messages.
    • No "simply" or "just" in front of a step; it reads as a judgement when the step is hard.

    Drifting when: Exclamation marks pile up, the cheerfulness carries into an error message, or filler like "we hope this helps" appears at the end of everything.

  3. Example 3

    Dry and witty

    Suits: Developer tools, newsletters and consumer brands whose audience already knows them.

    We areWe are not
    Wittyjokey
    Irreverent about the workirreverent about the customer
    Briefcryptic
    “Scheduled reports now come as PDFs, for everyone whose boss still prints things. Choose PDF in the schedule settings.”

    Rules that produce it

    • At most one joke per piece, aimed at the situation, never the reader.
    • Delete the joke and the message must still be complete.
    • No puns in headings or buttons.
    • Humour off for outages, errors, billing and security.

    Drifting when: The joke arrives before the information, slang that will date in a year, or sarcasm that lands on the customer.

  4. Example 4

    Precise and technical

    Suits: APIs, infrastructure, healthcare, legal and other regulated or technical audiences.

    We areWe are not
    Exactdense
    Formalstiff
    Completeexhaustive
    “Scheduled reports support PDF output. Select PDF under Delivery in the schedule settings. Existing schedules keep their current format until it is changed.”

    Rules that produce it

    • Use the product's exact interface labels and one term per concept.
    • State defaults and what does not change.
    • No contractions and no figurative language.
    • Numbers as numerals, with units.

    Drifting when: Synonyms creep in for defined terms ("export", "download" and "send" for one action), or vague timing like "soon" replaces a date.

  5. Example 5

    Bold challenger

    Suits: Brands entering a category against incumbents, launches and campaigns.

    We areWe are not
    Opinionatedcombative
    Confidentarrogant
    Punchyshouty
    “Stop exporting reports by hand. Scheduled reports now arrive as PDFs, ready before the meeting starts.”

    Rules that produce it

    • Short declarative sentences; imperatives are allowed in headlines.
    • Name the old way of working, not a competitor.
    • No capitals for emphasis and no superlative the brand cannot prove ("best", "fastest", "#1").

    Drifting when: Claims nobody can check, digs at named competitors, or every sentence written as a slogan.

  6. Example 6

    Calm and reassuring

    Suits: Finance, insurance, healthcare, customer support and incident updates.

    We areWe are not
    Calmcold
    Clearclinical
    Reassuringdismissive
    “You can now receive scheduled reports as PDFs. Your existing schedules stay as they are unless you choose PDF in the delivery settings.”

    Rules that produce it

    • Say what stays the same alongside what changes.
    • One idea per sentence.
    • No urgency words ("act now", "immediately") unless the reader really must act.
    • No humour.

    Drifting when: Hedging that buries the point, apologies for things that went fine, or urgency that was never needed.

Turn an example into a spec an AI tool can follow

A sample paragraph shows a model what the voice sounds like; it does not tell it why. Pair every sample with the rules that produced it, and write each rule as a decision: when it applies, what to do, and what would count as breaking it. Here is the plain-spoken expert voice written that way:

voice: plain-spoken-expert
traits:
  - direct, not curt
  - knowledgeable, not lecturing
  - confident, not boastful
rules:
  - when: any announcement
    do: state the change in the first sentence
    broken_if: the first sentence is about the team or its excitement
  - when: any sentence
    do: one idea per sentence
    broken_if: a sentence joins two instructions with "and"
  - when: describing the product
    avoid: [very, incredibly, game-changing, revolutionary]
tone_shifts:
  errors: drop warmth words, lead with what happened and what to do
examples:
  - "Scheduled reports now export to PDF. Pick the format when you set the schedule."

The field names do not matter; the structure does. Keep the spec in one place every tool reads from, rather than pasted into each prompt, so a change reaches every writer at once. Brand Voice Tokens for LLM Prompts and Few-Shot Example Selection for On-Brand LLM Output go further on which rules and samples to include, and Translating Legacy Brand Guidelines Into Structured Voice Specs covers converting a guide you already have.

Check your own voice with the swap test

Put a competitor's name in place of yours in your voice guide. If nothing else needs to change, the guide describes your category, not your brand, and any AI tool reading it will sound like your category too.

Then run the same check on output: take a draft, hide the logo and ask whether someone who knows your brand would recognise it. Evaluating LLM Output for Voice Fidelity turns that question into checks you can repeat.

Where to read real brands' voice guides

Some organisations publish their voice and style guides in full, and they are worth reading for the rules rather than the adjectives. Two well-known public ones are Mailchimp's Content Style Guide and the GOV.UK style guide. To find more, search for a company's name with "content style guide" or "voice and tone"; what each one contains, and whether it is still current, is best checked on the guide itself.

Questions

What is a brand voice example?

A short piece of writing that shows how a brand sounds, ideally next to the traits it demonstrates and the rules that produced it. On its own, a sample shows the result; with the rules, a writer or an AI tool can reproduce it.

How many traits should a brand voice have?

Few enough that a writer can keep all of them in mind while drafting. Each trait needs its own rules, samples and drift checks, so every trait you add is more to maintain and more for a model to balance.

What is the difference between brand voice and tone?

Voice is the brand's constant character. Tone is how that character adjusts to the situation, such as an announcement versus an error message. A good voice guide defines both.

Can an AI tool learn our brand voice from examples alone?

Examples help, but a model copies their surface: length, punctuation and topics. Pair them with explicit rules and the drift you want caught, and keep all of it in one shared source every tool reads.