Skip to content

Codify Brand Voice, Banned Phrases, and Tone Rules for AI Drafting

Eliminate generic robotic phrasing across team emails, proposals, and posts by turning vague brand adjectives into testable syntax rules and before/after pairs.

Copy the prompt and paste it into the AI assistant you already use.

How to make AI drafts sound like your company, not a bot

  1. 01

    Replace abstract adjectives with structural rules

    Swap 'sound authoritative and warm' for measurable rules: active verbs, 14–20 word average sentences, concrete numbers in paragraph one, zero exclamation marks.

  2. 02

    Maintain an explicit banned-phrase blocklist

    Forbid telltale filler ('in today's fast-paced landscape', 'game-changing', 'seamless', 'tapestries', 'not just X, but Y') directly in the system instructions.

  3. 03

    Anchor with three real gold-standard excerpts

    Include three short 80-word paragraphs written by your sharpest human communicator—few-shot examples steer style 5x better than rules alone.

  4. 04

    Include a bad-draft -> rewritten calibration pair

    Show one stiff generic draft alongside your exact edited company version so the model sees what you cut and tighten.

Common questions

Why do traditional PDF brand books fail when pasted straight into AI tools?

Agency brand decks focus on visual moodboards and aspirational adjectives, giving language models zero concrete guidance on cadence, vocabulary, or banned syntax.

How long should a reusable company brand-voice system prompt be?

Keep core voice instructions and three short calibration snippets between 400 and 700 words so they leave ample context budget for task facts.

Should customer support and outbound sales share the exact same voice prompt?

Share one 250-word company foundation block (banned cliches, terminology, honesty rules) paired with lightweight 150-word channel add-ons for support empathy or executive brevity.

Learn more

Related guides and background articles for this task