The Zitably Score: 7 Axes That Predict AI Citations

The Zitably Score: 7 Axes That Predict AI Citations

Most sites score below 40/100. That's not a judgment — it's a measurement of how well AI systems can parse, understand, and cite your content. Here's what the score means and how to improve it.

Two-Phase Architecture

The Zitably Score combines mechanical analysis (deterministic, instant) with LLM-powered evaluation (contextual, deeper). This two-phase approach ensures that the free tier provides real value while the Pro tier adds AI-powered insights.

Phase 1: Mechanical (/50 points) — Runs locally, zero LLM cost, <50ms per page.

  • Structure: /15
  • Extractability: /15
  • Technical Signals: /10
  • AI Accessibility: /10

Phase 2: LLM (/50 points) — Requires Pro tier, uses Gemini Flash, ~3–5s per page.

  • Answerability: /20
  • Fact Density: /15
  • Information Gain: /15

Combined: /100 → Grade A through F

Axis 1: Structure (/15)

What it measures: Heading hierarchy correctness, logical content flow, section organization.

How it's scored:

  • Single H1 present and descriptive (+3)
  • H2/H3 hierarchy without skipped levels (+4)
  • Logical section progression (intro → detail → conclusion) (+4)
  • Lists and tables used where appropriate (+4)

Common failures:

  • Multiple H1 tags (or none)
  • Jumping from H1 to H4
  • Wall-of-text paragraphs without structural breaks
  • 2,000-word pages with only one heading

How to fix: Use one H1 for the page title. Break content into sections with H2 headings. Use H3 for subsections. A section should cover one topic and be 150–400 words.

Axis 2: Answerability (/20)

What it measures: Whether the content provides direct answers to likely questions in its opening sentences.

How it's scored (LLM-evaluated):

  • Section openings directly answer the heading's implied question (+8)
  • Definitions and explanations lead with the answer, not background (+6)
  • Content addresses "what," "how," and "why" questions explicitly (+6)

Why it matters: AI retrieval systems rank content that leads with answers. If your section titled "What Is GEO?" starts with three paragraphs of history before defining GEO, the AI might not extract the definition.

How to fix: For every section heading, make the first sentence a direct answer to the question the heading implies. Then expand with detail.

Axis 3: Extractability (/15)

What it measures: Whether AI systems can pull discrete facts without ambiguity.

How it's scored:

  • Facts in bulleted/numbered lists (+4)
  • Data in proper HTML tables (+4)
  • Clear delimiter between facts (not buried in prose) (+4)
  • Consistent formatting patterns (+3)

Common failures:

  • Key data points buried in paragraph 7 of a long section
  • Tables rendered as divs instead of <table> elements
  • Statistics mixed with opinions without clear separation

How to fix: If a piece of information is a fact (number, date, name, specification), put it in a list or table. Don't hide data inside narrative prose.

Axis 4: Fact Density (/15)

What it measures: The ratio of verifiable, concrete information to filler text.

How it's scored (LLM-evaluated):

  • Named entities per 500 words (companies, people, products) (+5)
  • Numbers and statistics per 500 words (+5)
  • Verifiable claims vs. vague assertions (+5)

The benchmark: GEO research shows that pages with 12+ concrete facts per 500 words are 3.4× more likely to be cited than qualitative text.

How to fix: Replace "many companies use our product" with "deployed by 450+ organizations including Siemens, Bosch, and BASF." Replace "very fast" with "sub-1.2ms read latency under 10,000 concurrent requests."

Axis 5: Information Gain (/15)

What it measures: Whether the content introduces unique data not available elsewhere.

How it's scored (LLM-evaluated):

  • Original research, surveys, or case studies (+6)
  • Expert quotes or first-party data (+5)
  • Unique analysis or methodology (+4)

Why it matters: AI reranking algorithms are designed to filter out generic content. If your page says the same thing as 50 others, the reranker will favor the source with unique data.

How to fix: Add original data. Run a survey. Include expert quotes. Share internal metrics. Anything that can't be found on a competitor's page.

Axis 6: Technical Signals (/10)

What it measures: Whether technical markup supports AI discovery and extraction.

How it's scored:

  • Schema.org JSON-LD present and valid (+4)
  • llms.txt file exists at domain root (+3)
  • Clean HTML5 semantic elements (article, section, nav) (+3)

How to fix: Add Article, FAQPage, or Product schema to your pages. Create an llms.txt file. Use semantic HTML instead of generic <div> wrappers.

Axis 7: AI Accessibility (/10)

What it measures: Whether the content is available in AI-optimized formats.

How it's scored:

  • Content negotiation active (serves Markdown on Accept: text/markdown) (+5)
  • <link rel="alternate" type="text/markdown"> in HTML head (+2)
  • Content parseable without JavaScript (+3)

How to fix: Install Zitably (handles content negotiation automatically) or implement Markdown serving manually. Ensure your content renders in HTML without requiring client-side JavaScript.

Grade Scale

Grade Score Meaning
A 85–100 Excellent — AI-optimized, high citation probability
B 70–84 Good — minor improvements needed
C 55–69 Average — significant gaps in AI readability
D 40–54 Below average — major structural issues
F 0–39 Poor — largely invisible to AI systems

Getting Your Score

  1. Free tier — Install the Zitably plugin. Get your mechanical score (/50) instantly for any page.
  2. Pro tier (€19/mo) — Get the full 7-axis score (/100) with LLM-powered analysis and actionable suggestions per axis.

The Pro analysis also returns specific, prioritized recommendations: "Add FAQ schema," "Lead section 3 with a direct answer," "Replace paragraph 2's vague claims with statistics."

Ready to score your site? Install Zitably → or learn what GEO is first →