Brand voice guidelines your AI actually follows.
Your voice survives your writers — they've read the guide. It rarely survives the model. Pecia turns voice guidelines into machine-readable rules and checks every AI draft against them, in whatever tool your team writes.
What is brand voice?
Brand voice is how your company sounds: the tone it takes, the words it chooses, the words it refuses. Visual identity makes you recognisable before anyone reads a line; voice keeps you recognisable inside the line. Customers notice when it slips — not consciously, but the way you notice a colleague having an off day.
Why voice guidelines fail at AI speed
Most brand voice guidelines are a PDF: tone attributes, some do/don't examples, a terminology table. Well made — and structurally unable to do the job now, because the writer has changed. Every AI prompt is a brand-new writer with no memory of your guide, and pasting "write it in our brand voice" into the prompt produces an approximation of an approximation.
Tone of voice guidelines that live in a document govern people. They don't govern models. Models need the rules as data.
From guidelines to rules a model can follow.
Pecia holds your voice the way it holds your palette — as exact, machine-readable values:
- Tone attributes — how you sound, and the failure modes you refuse
- Terminology — the approved names for your products, and the banned alternatives
- Phrasing examples — real do/don't pairs — models learn fastest from contrast
- Address — how you speak to the reader, first person or second, formal or plain
one rule set · every AI tool
That structure doubles as a working brand voice guide template: if you can fill those four sections, you have a voice guide an AI can actually follow — whether or not you use Pecia to enforce it.
Tone of voice, enforced like a palette.
Connected over MCP, your AI tools query these rules the way they query your colours — automatically, on every draft. Messaging and terminology are checked against your voice before the send, the post, the page goes out; drift is corrected in the flow of work, not flagged in a review three days later. The same pass that keeps the logo inside its safe area keeps the copy inside your voice. One system, both halves of your identity — how the visual half works.
One voice, every AI tool
An AI brand voice tool that only works in one editor re-creates the original problem — your voice holds in one tool and drifts in four others. Pecia is model-agnostic: define the voice once and ChatGPT, Claude, Gemini, Copilot and any MCP-compatible client inherit the same rules. Swap models next quarter; the voice holds. New tool in the stack; it inherits the rules on day one. Connecting a tool takes one JSON snippet, and the same layer covers the decks your team generates too.
Questions.
- What is brand voice?
- Brand voice is the consistent personality in your company's writing — the tone, vocabulary and phrasing that make your content recognisably yours, whoever (or whatever) wrote it.
- What belongs in brand voice guidelines?
- Tone attributes, approved and banned terminology, real do/don't phrasing examples, and how you address the reader. Written for humans, that's a guide; structured as data, it's a rule set an AI can follow.
- How do you maintain brand voice with generative AI?
- Stop re-explaining the voice in every prompt. Make the guidelines machine-readable and let the AI query them automatically — that's what Pecia's MCP server does — so every draft starts from your rules, not from the model's average.
- How do you adapt AI-written text to your brand voice?
- Editing after the fact works once; it doesn't scale. The scalable version is enforcement at generation: the draft is checked against your tone and terminology rules and corrected before it leaves the tool.
- How do agencies maintain brand voice with AI?
- The same way in-house teams do, multiplied by every client: one machine-readable rule set per brand. With the rules served over MCP, whichever AI tool an account team uses answers in the right client's voice.
- How do you keep a consistent brand voice in personalized content?
- Personalisation changes the message, not the voice. Hold the variables — name, segment, offer — and enforce the constants: tone, terminology, phrasing. Rules as data make that separation checkable on every variant.
