Brand Voice Rewriter
Extract a voice from samples you like, then rewrite copy into it without changing facts.
Builds a working voice definition from writing samples you already like, then rewrites supplied copy line by line into that voice, showing each change and flagging any rewrite that altered meaning instead of tone.
Ready-to-use prompt
The prompt
Copy it as-is, then swap the bracketed placeholders for your own details before running it.
Role: You are a brand voice editor.
Context:
- Three to five samples of writing that sound right: {{voice_samples}}
- Samples that sound wrong, if you have them: {{anti_samples}}
- Copy to rewrite: {{copy}}
- Audience: {{audience}}
- Non-negotiables, legal lines, product names, claims: {{fixed_elements}}
Task: In two parts.
Part one, the voice definition:
1. Six traits, each with a do and a do-not example taken from the samples
2. Sentence rhythm: typical length, variation, use of fragments
3. Vocabulary rules: words to use, words to avoid, contractions, jargon level
4. How the voice handles bad news, uncertainty and selling
5. What this voice is not, using the anti-samples
Part two, the rewrite:
- Rewrite the supplied copy line by line
- For each changed line show BEFORE, AFTER and the trait applied
- Leave every fixed element untouched and list them
- Flag any line where the rewrite changed the meaning, not just the tone
Rules:
- Derive the voice from the samples only, do not import generic brand adjectives.
- Never alter a factual claim, a number or a legal line while restyling.
- If a line already fits the voice, leave it and say so.
- If the samples do not establish a rule, mark it as undetermined rather than filling the gap.
- Preserve source order and assign each line a stable identifier.
- For each meaning-change flag, provide a meaning-preserving alternative or retain the original.
Finish with: the three rules a new writer would need to sound like this brand tomorrow.Estimated results
Editor's note
Why this prompt matters
Brand voice guidelines built from adjectives — confident, human, bold — are unusable, because three writers will read them three ways and all three will be defensible. Voice extracted from real samples is a different artifact: it contains sentence rhythm, actual vocabulary rules, and a description of how the brand behaves when it has bad news. This prompt produces that definition first and only then rewrites, which is why the definition is worth saving separately and pasting into every future prompt you run.
Anatomy
Prompt engineering breakdown
Role
Context
Goal
Constraints
Output format
What you'll get
Expected output
Scenario and source material
A hypothetical file-sharing service needs a permission notice rewritten. The audience is someone who has just lost sharing access. These invented examples illustrate the workflow; they are not verified product documentation.
Approved samples
- “Your saved files are unchanged. You can still open them.”
- “Sharing is paused. Ask the folder owner to restore access.”
- “You may be able to restore sharing after the owner reviews your request. Try a shared folder if your team needs one place for files.”
Anti-samples
- “Oops! Sharing hit a tiny snag. Unlock seamless teamwork now!”
- “Restoration is guaranteed. Initiate the permissions remediation process.”
Fixed element: “Your saved files are unchanged.”
Part one: working voice definition
These traits describe observable choices in the supplied text, not an aspirational brand personality.
| Trait | Do: source example | Do not: source example | |---|---|---| | Calm about disruption | “Sharing is paused.” | “Oops!” | | Literal about status | “Sharing is paused.” | “Sharing hit a tiny snag.” | | Specific about the next step | “Ask the folder owner to restore access.” | “Unlock seamless teamwork now!” | | Familiar vocabulary | “restore access” | “permissions remediation process” | | Honest about uncertainty | “You may be able to restore sharing” | “Restoration is guaranteed.” | | Conditional when selling | “if your team needs one place for files” | “Unlock seamless teamwork now!” |
Rhythm: Mostly short, complete sentences, with a longer sentence when a condition matters. The approved samples do not establish a preference for fragments.
Vocabulary: Prefer “sharing,” “files,” “ask,” and “owner.” Avoid metaphorical “unlock,” promotional “seamless,” and administrative jargon. Contractions are not evidenced; do not infer a ban or preference.
Difficult moments: State bad news without jokes or minimising it. Preserve uncertainty explicitly. Present a product suggestion only alongside the need it serves.
Part two: line-by-line rewrite
| Before | After | Trait applied | |---|---|---| | Unfortunately, sharing functionality is currently paused. | Sharing is paused. | Calm; familiar vocabulary | | Your saved files are unchanged. | Your saved files are unchanged. | Already fits; fixed element | | Please contact the folder owner to request restoration of sharing access. | Ask the folder owner to restore sharing access. | Specific next step | | Sharing may become available again following review by the owner. | Sharing may become available again after the owner reviews your request. | Familiar vocabulary |
Meaning check: The last rewrite assumes the review concerns a request. That is plausible but not explicit in the original. Flag it for confirmation; use “after the owner's review” instead if unconfirmed. No guarantee of restoration should be introduced.
Fixed elements retained: “Your saved files are unchanged.”
Three rules for the next writer
- Name the situation before giving the next step.
- Replace jargon, not conditions or qualifications.
- Offer help without jokes, hype, or guaranteed outcomes.
Under the hood
Why this prompt works
Deriving traits from samples rather than adjectives is what makes the definition enforceable: a new writer can compare their draft against a do and a do-not example and settle the question themselves. The meaning-drift flag handles the genuine risk in any restyling pass — a model tightening a sentence will happily drop a qualifier that was doing legal or factual work, and the change reads like an improvement. Anti-samples matter more than most teams expect, because a boundary defines a voice faster than a description does.
Model fit
Best AI models for this prompt
Claude
Best here. Most faithful to a sample voice and the most reliable at flagging meaning drift.
ChatGPT
Fast on long rewrites; restate the do-not-change-facts rule for safety.
Gemini
Handles many long samples in a single pass when building the definition.
When to use
- Before a freelancer's first assignment, to create an evidence-backed voice brief.
- When combining several authors' onboarding messages into one sequence.
- During a rebrand migration, after approving representative new-voice samples.
- Before publishing a revised support notice, to inspect tone changes beside the original.
When not to use
- When nobody can identify approved samples; resolve the editorial direction first.
- For protected legal wording; have counsel approve any proposed changes separately.
- When claims are disputed; consult product documentation or the responsible owner first.
- For sensitive incident messaging where a human must judge accountability, empathy, and disclosure.
Get more from it
Pro tips
- 1
Supply actual passages, not adjectives such as “friendly”; require each trait to cite wording that demonstrates it.
- 2
Pair anti-samples with similar situations so the boundary reflects voice, not subject matter.
- 3
Review meaning flags against the source, especially permissions, promises, deadlines, and implied guarantees.
- 4
Save the approved voice definition with its supporting samples for reuse; discard unsupported traits first.
- 5
Put exact protected strings in {{fixed_elements}}, including punctuation where preservation matters.
- 6
Keep original line breaks and assign line numbers before submitting long copy; compare unchanged lines too.
- 7
For a second pass, identify failed lines and the violated trait. Request targeted corrections, not another wholesale rewrite.
Don't ship this
Common mistakes
✗ Defining voice with adjectives.
Fix — Extract traits from real samples with do and do-not examples.
✗ Letting a rewrite alter a number or a claim.
Fix — List fixed elements and review the meaning-drift flags.
✗ Restyling copy that is factually wrong.
Fix — Fix the substance first; voice cannot rescue a bad argument.
People also ask
Frequently asked questions
Q.Why does it need writing samples?
Because voice defined by adjectives is unenforceable. Traits extracted from real samples come with do and do-not examples, so a new writer can settle a wording question without asking anyone.
Q.What is the meaning-drift flag?
A warning that a rewrite changed substance rather than style. Models tightening a sentence often drop a qualifier that was doing factual or legal work, and the result reads like an improvement until someone checks.
Q.Can I reuse the voice definition elsewhere?
Yes, and you should. Save part one and paste it as the voice block into every other copy prompt you run, so output arrives on-brand instead of needing a separate editing pass.