Brand Voice Guidelines for Writers and AI Tools
Brand voice guidelines are the written rules for how your company sounds, so every writer and AI tool writes the same way. Eicra BD in Dhaka writes the principles, tone map, word list and worked examples. We then turn them into prompt files your team loads into its AI tools. A strategist, a senior writer and a reviewer work on each project.
One exit at each step. No retainer until you ask for one. Scoping is free. The diagnostic gives you a written verdict on how far your published content drifts from one voice. You can stop there, take the findings to another supplier, or move to the pilot. Nothing we write is licensed back to us.
You keep the verdict either way.
One call 01 Scope · free
5 working days 02 Diagnose · $99, credited to the pilot
2 weeks 03 Pilot · $1,200, one channel
4–6 weeks 04 Production · from $3,600
Monthly 05 Managed Ops · from $999 a month, optional
What is included in brand voice guidelines?
Brand voice guidelines are the written rules for how your company sounds on every channel: three to five voice principles built on your brand’s personality traits, a tone map for different situations, a word list of terms to use and avoid, rewritten before-and-after examples, and prompt files so AI tools produce copy in the same voice.
Lists of the 5 C’s of branding differ by source. Most include clarity and consistency, and many add credibility or connection; the rest vary, from confidence to commitment. Brand voice guidelines deliver the first two directly: one clear way of writing, used the same way on every channel.
Voice principles
Three to five named principles, each with a plain rule, the reason behind it and one short line of copy showing the principle applied. Each principle also says what it does not mean, so writers do not stretch it. Principles are built from your brand’s personality traits, not from a template.
Tone map
One grid showing how the voice shifts across sales pages, support replies, error messages, social posts and apologies, with a sample sentence for each. Writers check the grid before they start, so the tone fits the moment. Support and sales teams use the same map, which keeps them sounding like one company.
Brand language guidelines
A working word list: terms you own, terms you avoid, spelling and capitalisation rules, product naming, and the grammar decisions writers argue about weekly. The list settles those arguments once, in writing. It is kept short enough to use, and new terms are added as products launch.
Brand messaging matrix
One page per audience persona holding the claim, the proof, the objection and the call to action: a messaging framework campaigns reuse. Writers and AI tools both start from the same claims. Every claim is tied to proof you can show, so nothing is invented for a campaign.
Before-and-after examples
Ten pieces of your own published copy rewritten in the new voice, side by side, so reviewers judge the change. Each rewrite notes which principle changed the copy. Your team sees the rules at work on its own content, not on a made-up sample.
AI brand voice prompt files
System prompts, a style rule file, few-shot examples and a checklist your team pastes into the models and writing tools it already runs. Each file is tested on your own briefs before handover. The files are plain text, so your team can edit them without coming back to us.
| Aspect | Brand voice | Tone of voice |
|---|---|---|
| What it is | The principles, vocabulary and point of view | How that voice adjusts to the situation |
| Does it change? | No: it stays the same everywhere | Yes: warmer in an apology, shorter in an error message |
| Where it is written | Voice principles and the word list | The tone map |
Not included: A production AI system or agent build, since this service delivers documents and prompt files · the monthly review of new and generated copy after handover, which is Managed Ops, optional and priced monthly · any retainer you have not asked for.
Where are tone of voice guidelines used?
Tone of voice guidelines earn their fee where copy is produced fastest. Most teams apply them first to the website, then to lifecycle email, support replies and AI-assisted drafts, because those four channels create most of the content a customer actually reads.
Website and landing pages
Headlines, sub-headings and calls to action rewritten against the principles before a page ships. Each page is read against the tone map before it goes live. Your team gets notes that point to the rule behind each change, so the same fix is not needed twice.
Lifecycle and sales email
Sequences checked so onboarding, renewal and win-back mail sound like one company, not four. Every email in a sequence is read in order, the way a customer gets it. Tone shifts are marked, and the word list flags terms that do not belong in your voice.
Support and product copy
Error messages, empty states and help replies carry the same tone even under pressure. These are the lines customers read when something has gone wrong. The tone map gives writers a sample for each case, so replies stay calm, short and clear.
AI-assisted content
Prompt files keep generated drafts inside the voice, with a reviewer signing each batch. Every batch is logged with its prompt, its output and the reviewer’s name. When a draft misses the voice, the prompt file is fixed, not just the draft.
How is an AI draft checked for voice?
One AI-drafted email is checked against brand voice guidelines in five logged steps: the brief loads the voice prompt file, the draft is generated inside the tone map, the word list check flags banned terms, a reviewer signs or returns the batch, and approved copy is published. The brand voice example below shows the chart and the check.
The guide · brand voice chart
Abc Ltd · voice guideFor writers and AI tools
We are · we are not
Clearnot blunt
Warmnot casual
Confidentnot loud
Helpfulnot pushy
Tone by channel · warmth
Sales pageconfident, specific
Support replywarm, patient
Error messageshort, plain
Apologywarmer, direct
Words to use
sign ininvoiceyour team
Words to avoid
world-classleveragesynergy
Applied · one renewal email, checked
Renewal email · batch of 12 draftsTone: warm, short
To: abc@xxx.com.bd · From: Abc Ltd
Your plan renews next month and nothing changes unless you ask. Our world-class support team is one reply away.
- 1Brief loaded with the voice prompt fileDone
- 2Draft generated inside the tone mapDone
- 3Word list check flags banned terms1 flag · world-class
- 4Reviewer signs or returns the batchA person signs
- 5Approved copy published and logged for reviewLogged
Managed Ops: a monthly review of new and generated copy updates the word list, the prompts and the examples.
Illustrative example. Yellow marks the step where a person signs; every AI-assisted batch is logged with its prompt, its output and the named reviewer.
How are brand voice guidelines developed?
Brand voice guidelines are developed in five steps: we scope the channels, diagnose the drift in your published copy, pilot the guide on one channel, produce the full system with prompt files, then keep it current as products and campaigns change. There is one exit at each step.
Lists of the 7 pillars of branding differ by source too. One common list names purpose, perception, identity, values, brand experience, positioning and promotion. Brand voice is the written side of identity. The guidelines turn that pillar into rules a writer or an AI tool can follow.
01 Free
Scope A free call to list the channels, the writers, the tools and the decisions that keep being argued about, with no document requested yet.
02 5 working days
Diagnose We read your published copy across channels and return a written verdict on where the voice breaks and what to fix first.
03 2 weeks
Pilot One channel gets the draft principles, tone map and word list, applied to real copy, so the guide is tested on actual work.
04 4–6 weeks
Production The full system is written and handed over with a working session: principles, tone map, language rules, messaging matrix, examples and prompt files.
05 Monthly
Managed Ops A monthly review of new and generated copy against the guide, with updates to the word list, the prompts and the examples.
| Step | Time | Price | You receive |
|---|---|---|---|
| Scope | One call | Free | Channels, writers, tools and open decisions listed |
| Diagnose | 5 working days | $99, credited to the pilot | A written verdict on where the voice breaks |
| Pilot | 2 weeks | $1,200, one channel | Draft principles, tone map and word list tested on real copy |
| Production | 4–6 weeks | From $3,600 | The full system, examples and prompt files, handed over |
| Managed Ops | Monthly, optional | From $999 a month | Review of new and generated copy, with updates |
Who writes your verbal identity system?
Each project runs with a named strategist, a senior writer and a reviewer, plus an AI engineer who turns the finished rules into prompt files your tools can load. You know each person by name before work starts. The reviewer is not the writer, so every batch gets a second pair of eyes.
Voice strategist Sets principles and the tone map
Senior writer Writes the rules, examples and rewrites
Reviewer Signs every batch against the guide
AI engineer Builds and tests the prompt files
Researcher Reads competitor, customer and support language
Account lead Runs the calls and the change log
Do you need a brand voice agency?
Not every company needs a brand voice agency on retainer. Answer the five questions below and the panel returns a verdict: a one-off guide, a pilot on one channel, a full verbal identity system, or a monthly review of AI-assisted content.
What is the 3-7-27 rule in branding? It is a marketing rule of thumb, not a research finding: a buyer needs about 3 exposures to notice a brand, 7 to remember it and 27 to trust it. A written voice keeps those 27 exposures sounding like one company, whoever writes them.
1. How many people write customer-facing copy?
2. How much of your copy starts in an AI tool?
3. Do you have written voice rules today?
4. How many channels publish in a normal month?
5. Who signs copy off before it ships?
Which level do you need?
Answer the five questions and the panel suggests the smallest fit. That may be the diagnostic alone, a pilot on one channel, the full verbal identity system, or a monthly review of AI-assisted copy. Nothing is sold before you have read the verdict.
Start with the diagnostic
Few writers and little AI drafting: buy the written verdict first, then decide after reading where your published copy actually breaks and what it costs. You can stop after the verdict, take it to another supplier, or move to a pilot. The $99 fee is credited to the pilot if you go ahead.
How the verdict is decided
Most drafts start in an AI tool → add managed review
Mostly first options → start with the diagnostic
Mixed answers → a pilot on one channel
Mostly third options → the full verbal identity system
Why choose Eicra BD for brand voice guidelines?
Most tone of voice services stop at a style guide; we treat the guide as an operating file that writers and language models load the same way, with a monthly review that keeps both inside the voice. A named reviewer signs every AI-assisted batch, and personal data follows Article 28 of the General Data Protection Regulation (GDPR).
Article 28(3)(b) GDPR says the processor “ensures that persons authorised to process the personal data have committed themselves to confidentiality or are under an appropriate statutory obligation of confidentiality”.
Without written voice rules
This is how most teams work before a guide exists. Each writer follows their own habits, reviews turn into arguments about taste, and AI drafts change voice from week to week. The cost shows up as rework on every piece.
!!!!!
- The voice lives only inside one senior writer’s head today
- Every reviewer argues from personal taste instead of a written rule
- AI drafts read like a different company each week
- Campaign claims are rewritten from memory every single quarter
With the guide in place
With written rules, a review checks copy against the guide instead of opinion. New writers learn the voice from the principles and examples, and prompt files keep AI drafts inside it. The illustrative batch review below shows how one check is recorded.
Batch review · renewal emails
Drafts12Word list flags3 fixedToneWarm, shortReviewerSignedVerdictPublished
Illustrative example
- Written principles any new writer can apply on day one
- Reviews are settled against a rule, not personal opinion
- Prompt files keep generated drafts inside the agreed voice
- The messaging matrix is reused across every campaign brief
Is outsourcing brand voice work offshore safe?
Outsourcing brand voice guidelines to an offshore team is safe when the controls are written down before any unpublished campaign, pricing line or customer research moves. As a Bangladesh-based company, we work under a non-disclosure agreement (NDA) and a processing agreement, and every AI-assisted draft is logged and signed through human-in-the-loop review. Reviewed by Mohammad Nazrul Islam, Managing Director
Which agreements are signed first?
Every agreement below is signed before any unpublished campaign, research file or pricing document is shared. That means an NDA, a processing agreement wherever personal data appears, and named accounts only. Certifications are listed only when they are held; none is claimed on this page.
NDA — signed before any unpublished campaign, research file or pricing document is shared with us.
Processing agreement — terms under Article 28(3) of the GDPR wherever personal data appears in your research files.
Access — named accounts only, with access removed within one working day of a person leaving.
Tools — we work inside your drive, your workspace and your writing tools where you prefer.
Logs — every AI-assisted batch is recorded with its prompt, its output and the named reviewer.
What ownership and rework terms apply?
Everything written for you stays yours. The guide, the examples and the prompt files are yours outright, in editable formats. Customer quotes are anonymised before use, a named reviewer signs every batch, and rework is free where delivered copy misses the agreed rules.
Ownership You own the guide, the examples, the prompt files and every source draft outright.
Personal fields Customer quotes are anonymised before they enter any prompt, example or published rule.
Named reviewer A person signs each batch, so no unreviewed generated copy ever reaches a customer.
Rework Written rework at no fee where delivered copy misses the rules we agreed.
Certifications Listed only when held; nothing is claimed on this page.
What proof do you get before you pay?
Before you pay for brand voice guidelines, you get evidence instead of promises: a written diagnostic verdict on where your published copy breaks, a pilot tested on one channel of real copy before the full system is written, and a named reviewer on every AI-assisted batch. You can stop after any of them.
2–3 weeks
To a first voice guide draft, not months: the draft principles, tone map and word list.
Every batch
Of AI-assisted copy is signed by a named reviewer before it reaches a customer.
Every file
The guide, the examples and every prompt file are yours outright, in editable formats.
Case studies: client results with numbers are added here as clients give permission to name them.
What do buyers ask about brand voice guidelines?
Buyers of brand voice guidelines ask nine questions before they commit. They cover what each step costs, how long it takes, what the guide needs for AI tools, how a voice is created, and how voice differs from tone. They also cover existing style guides, which AI tools are supported, customer data, and what is outside the scope.
How much do brand voice guidelines cost?
The diagnostic costs $99, credited to the pilot. The pilot on one channel is $1,200 fixed, the full system starts from $3,600, and Managed Ops is optional from $999 a month. Each price is agreed before work starts, never as an hourly rate, and scoping is free.
How long does brand voice development take?
The full system, including the messaging matrix, worked examples and prompt files, usually lands in four to six weeks, depending on how many channels and reviewers are involved. The diagnostic takes about five working days, and a pilot on one channel runs two weeks.
What should a guide include for AI?
For AI tools, brand voice guidelines need machine-readable parts: a system prompt carrying the principles, a rule file of banned and preferred terms, worked before-and-after examples the model can imitate, and a reviewer checklist, because a document written only for humans cannot be loaded.
How do you create a brand voice?
Start with the brand personality framework: the traits and the point of view that justify each voice principle. Then read your published copy to find where the voice breaks, write three to five principles, a tone map and a word list, test them on one channel, and turn the result into examples and prompt files.
How is brand voice different from tone?
Voice is constant: the principles, vocabulary and point of view that stay the same everywhere. Tone shifts with the situation, so the same voice sounds warmer in an apology and shorter in an error message. The tone map records both in one grid.
Can we start from our style guide?
Yes. The diagnostic reads your current style guide next to your published copy and marks what to keep, what to rewrite and what is missing. Rules that already work stay in place, and the pilot builds on them instead of starting again. You read the written verdict before any pilot is priced.
Which AI tools are supported?
The prompt files are written for the AI tools your team already uses, most often ChatGPT, Claude, Gemini or Microsoft Copilot. Each file is tested on your own briefs in those tools before handover. The files are plain text, so they also load into writing tools that accept custom instructions.
How is our customer data protected?
Work starts under an NDA and, where personal data appears, a processing agreement under Article 28(3) of the GDPR. Customer quotes are anonymised before they enter any prompt, example or rule. We work inside your drive and tools where you prefer, with named accounts only, and access is removed within one working day of a person leaving.
What is outside the scope?
This service covers written voice only. Logo, colour and visual identity design are not included, and neither is a rewrite of your full website or content library. Prompt systems for whole content workflows are covered by prompt engineering for marketing. We do not build AI software or agents.
Get a written verdict on where your brand voice breaks.