Brand Voice

Brand Voice & Tone Analysis

This skill teaches Claude how to analyse brand voice and content strategy from scraped website text. The analysis produces a structured voice profile usable for content creation and brand guidelines.

Analysis Framework

Step 1: Content Collection

Categorise all scraped text into:

• Headings (h1–h6): Brand messaging hierarchy
• Body copy: Communication style and complexity
• CTAs: Action language patterns
• Navigation labels: Information architecture language
• Form labels & placeholders: Instructional tone
• Footer content: Legal/formal register
• Error messages & empty states: Empathy and helpfulness
• Microcopy: Tooltips, badges, status text

Step 2: Tone Dimension Analysis

Rate each dimension on a 1–10 spectrum with evidence:

Each rating MUST include:

• The numeric score
• 1–2 specific evidence quotes (each under 14 words)
• Justification for the score

Step 3: Voice Characteristics

Identify 3–5 defining voice traits. Each trait needs:

• Trait name (adjective)
• Definition (one sentence)
• Evidence (specific quote from the site, under 14 words)
• Counter-example (what this brand would NOT say)

Example:

Trait: Confident
Definition: States capabilities directly without hedging or qualifying.
Evidence: "The fastest way to build financial infrastructure"
Counter-example: Would NOT say "We think we might be able to help with..."

Step 4: Language Variant Detection

Identify Australian, American, or British English:

Evidence must cite specific words found on the site.

Step 5: CTA Pattern Analysis

Collect all CTAs (button text, link text for actions) and analyse:

• Verb usage: Start with verb? Which verbs? (Get, Start, Try, Learn, Explore, Build, Join)
• Personalisation: "your" vs. generic ("Start your trial" vs. "Start trial")
• Length: Word count pattern
• Urgency: Time-limited language? ("Now", "Today", "Free")
• Specificity: Vague ("Learn more") vs. specific ("See pricing plans")

Document ≥3 CTA examples with pattern categorisation.

Step 6: Content Guidelines Generation

Produce at least 5 "do" and 5 "don't" guidelines. Each must be:

• Specific (not "be clear" but "use sentences under 20 words for feature descriptions")
• Evidenced (derived from actual patterns observed)
• Actionable (a content writer can follow it immediately)

Example:

DO: Lead CTAs with action verbs ("Start building", "Get started", "Explore features")
DON'T: Use passive CTAs ("Click here", "Submit", "More info")
Evidence: 8/10 observed CTAs begin with an active verb.

Output Format

{
  "tone_dimensions": [
    {
      "dimension": "Formality",
      "score": 4,
      "spectrum": "casual ↔ formal",
      "evidence": ["Direct, conversational headings", "Uses contractions throughout"],
      "justification": "Consistent use of 'you' and contractions suggests accessible, peer-level tone"
    }
  ],
  "voice_characteristics": [
    {
      "trait": "Confident",
      "definition": "States capabilities directly without hedging",
      "evidence": "The fastest way to build financial infrastructure",
      "counter_example": "We think we might be able to help"
    }
  ],
  "language_variant": {
    "detected": "American English",
    "confidence": "HIGH",
    "evidence": ["'color' spelling in UI", "'center' in layout text", "USD currency first"]
  },
  "cta_patterns": [
    {
      "text": "Start building",
      "category": "action-verb-lead",
      "verb": "Start",
      "personalised": false,
      "word_count": 2
    }
  ],
  "content_guidelines": {
    "do": ["Lead CTAs with active verbs", "..."],
    "dont": ["Use passive CTA language", "..."]
  },
  "vocabulary": {
    "preferred_terms": ["build", "scale", "infrastructure"],
    "avoided_terms": [],
    "industry_jargon": ["API", "SDK", "webhook"]
  }
}

Validation Criteria (Gate 4 — Voice)

• S-VOI-01: ≥4 tone dimensions rated with evidence
• S-VOI-02: 3–5 voice traits defined with evidence under 14 words each
• S-VOI-03: Language variant detected with evidence
• S-VOI-04: ≥3 CTA examples documented with pattern analysis
• S-VOI-05: ≥5 do's and ≥5 don'ts generated

Converted and distributed by TomeVault — claim your Tome and manage your conversions.