AI — Brand Voice Analysis

How AI analyzes
brand voice and personality.

ExtractVibe uses OpenRouter-connected language models to read website copy and classify tone, identify personality traits, and generate actionable writing guidelines. Here is how it works under the hood.

Get started free View docs

01 / AI — Brand Voice Analysis

Text Extraction

After the page is rendered, all visible text content is extracted from headings, paragraphs, buttons, navigation, and meta descriptions. Non-content text (scripts, styles) is filtered out.

02 / AI — Brand Voice Analysis

Tone Classification

The AI model classifies brand tone across multiple dimensions: formal vs. casual, serious vs. playful, technical vs. accessible, and authoritative vs. conversational. Each dimension is scored on a scale.

03 / AI — Brand Voice Analysis

Personality Profiling

Using a multi-trait model inspired by brand archetypes, the AI identifies dominant personality traits like innovative, trustworthy, bold, friendly, sophisticated, or rebellious.

04 / AI — Brand Voice Analysis

Guideline Generation

The AI produces concrete writing guidelines including sentence length recommendations, vocabulary level, punctuation patterns, and a list of do's and don'ts for maintaining brand voice consistency.

How it works

Text is extracted

All visible copy is scraped from the rendered DOM, including headings, body text, CTAs, and meta content.

AI analyzes patterns

Gemini Flash 2.0 analyzes word choice, sentence structure, emotional tone, and communication patterns.

Guidelines are generated

The model produces a structured voice profile with tone scores, personality traits, and writing do's/don'ts.

Code Example

Voice analysis API response

{
  "voice": {
    "tone": {
      "formality": 0.7,
      "enthusiasm": 0.6,
      "technicality": 0.8,
      "warmth": 0.5
    },
    "personality": [
      "innovative",
      "confident",
      "precise",
      "developer-focused"
    ],
    "writingStyle": {
      "sentenceLength": "medium",
      "vocabulary": "technical but accessible",
      "punctuation": "minimal exclamation marks"
    },
    "dos": [
      "Use active voice",
      "Lead with benefits",
      "Keep sentences under 20 words",
      "Use specific numbers and metrics"
    ],
    "donts": [
      "Avoid buzzwords and jargon",
      "Don't use excessive exclamation marks",
      "Avoid passive constructions",
      "Don't start sentences with 'We'"
    ]
  }
}

Frequently Asked Questions

Which AI model powers voice analysis?+

ExtractVibe uses Gemini Flash 2.0 via OpenRouter for voice analysis. This model offers strong language understanding at a low cost per token, making it ideal for analyzing brand copy.

How accurate is the voice analysis?+

Voice analysis is most accurate on pages with substantial copy (marketing pages, about pages, blog posts). Minimal pages with little text may produce less detailed results.

Can I customize the voice analysis dimensions?+

The API returns a fixed set of tone dimensions and personality traits. If you need custom dimensions, you can self-host ExtractVibe and modify the analysis prompts.