How to write brand voice guidelines AI tools can follow

Start with a rule a writer can use

A voice rule should settle a writing decision. For each one, record:

Rule and source:
Messages it applies to:
Reason:
Approved copy showing the decision:
Earlier draft and the edit it needed:
Facts the task must supply before writing:

If you cannot find an example or explain the reason, keep the rule as a proposal while you gather evidence.

For a Teenage Engineering newsletter, Rainbrand told ChatGPT to introduce the electronic instrument maker's OP–XY to OP-1 field owners without calling it an upgrade or replacement. The draft expressed that distinction in these lines:

meet OP–XY. think in sequences.

same spirit of exploration. a different way to make music.

Gather the copy and edits that define the voice

Gather writing you would use again:

• Include the formats you publish, their earlier drafts and the reasons for edits.
• Record each sample's source, format and approval status. Label new examples as proposals.
• Separate style edits from factual corrections, such as removing a claim about an unshipped feature.

Ask an agent to find the recurring decisions:

Read these approved writing samples and their earlier drafts. Group the repeated edits into proposed voice rules. For each rule, cite the sample and edit that support it, explain when it applies, and identify any conflicting example. Keep factual corrections separate from style preferences. Label every new rule or example as proposed.

Write rules that settle an edit

A trait needs an instruction beneath it. Adapt these to decisions your editors would enforce:

Tone by situation

Mailchimp distinguishes a relatively consistent voice from changing tone. Nielsen Norman Group's four tone dimensions are formality, humour, respectfulness and enthusiasm. Translate those choices into situations the agent can identify:

Choose the rows your work needs and add approved examples.

Vocabulary and sentence rules

List exact product spellings, preferred words and exceptions:

Add spelling, contraction, case and punctuation rules where useful. Treat length targets as prompts to inspect the writing; preserve product names and essential conditions.

Pair each rule with examples

OpenAI and Anthropic recommend varied examples close to the task.

For each pair:

1. Keep an actual draft beside its approved revision.
2. Explain the specific edit, such as moving an error's next action before a joke.
3. Distinguish style findings from factual corrections; future tasks need their own current facts.

Use plausible near-misses from different subjects and formats. See the published voice-guide examples for more ways to explain a decision.

Assemble the guide and the task request

Fill this worksheet, repeating the rule and example records as needed:

# [Brand] voice guide

## Audience and purpose

- Who we write for:
- What they need from our writing:
- Formats this guide covers:
- Source links and revision date:

## Voice rules

### [Trait]

- Rule:
- Applies when:
- Reason:
- If rules conflict, prioritise:
- Approved example and source:
- Near-miss and the edit it needs:

## Tone by situation

| Situation   | What changes in the writing | Facts needed |
| ----------- | --------------------------- | ------------ |
| [Situation] | [Practical instruction]     | [Task input] |

## Vocabulary and mechanics

- Exact names and spelling:
- Preferred words, replacements and exceptions:
- Spelling convention and contractions:
- Headings, buttons and punctuation:
- Length or structure rules, where useful:

## Examples by format

### [Format and example]

- Status: [approved copy / proposed edit]
- Source and date:
- Task facts used in this example:
- Copy:
- Rules illustrated:
- Near-miss and why it fails:

## Related factual sources

- Product capabilities and availability:
- Prices and terms:
- Supported claims:

## Revision record

- Rule changed and why:
- Failed writing case to rerun:

Link changing product facts to their own maintained sources. Each request should say what to make, who will read it and which current facts to use:

Write an outage email for our affected customers, using our
current voice guide and its outage tone rules.

Use the attached incident record for the affected period, impact,
current status and recovery steps. Do not reuse incident details from
the guide's examples. Identify any missing fact before drafting the
sentence that needs it.

Test the guide on the work it is meant for

Save representative tasks, including ordinary messages and difficult cases such as an apology or uncertain cause. This procedure applies Anthropic's task-specific evaluation guidance to voice:

1. Run the same tasks with and without the guide on your models, keeping facts and formats the same.
2. Check exact spelling, phrases and punctuation mechanically where useful.
3. Have a separate reviewer apply a written rubric. Compare some findings with edits from someone who knows the voice, especially when the rubric is new or reviewers disagree.
4. Save failures with the task, draft, rule and preferred revision. Identify missing facts, unclear guidance or a failure to follow a clear rule.
5. Correct the relevant input and rerun the tasks. Correct existing drafts separately.

Use this grader request:

Check this draft against the attached voice guide. For each applicable rule, quote the relevant passage, explain the finding, and mark pass, fail or not applicable. Keep voice findings separate from factual questions. Flag facts that need a source. Do not rewrite the draft. End with the specific edits needed.

Rerun the set after a guide or model change. Record which model and guide revision produced each draft.

Make the revised guidance available to agents

Rainbrand maintains brand context for connected agents and products:

• New brand: add the guide with the brand's sources and wait until the brand is ready.
• Existing brand: prepare an edit, inspect and refine the proposal, then apply or discard it.

Once the brand is ready, or after applying an update, request a new Brief, the guidance and files for the task. Include its audience, channel and situation.

Request a Brief for [brand] for the writing task attached here. Use its audience, channel and situation. Take facts from the linked current sources. Read the completed Brief, then draft using its voice guidance. Show the draft and identify any facts that remain unresolved.

Use the text grader above to check the draft's voice. Rainbrand Review can check a composed email, page or other visual deliverable from an image or capture; use the agent's tools to make corrections.

Update copies saved in other tools too. The shared-guidelines guide covers keeping those current.