The 12 Factors That Determine Your AI Visibility Score
Every AI visibility tool hides how they score. We publish ours. Here are the exact 12 factors, weights, and thresholds behind every Adymus score.
Ben Nawin
Most AI visibility tools hide how they calculate scores. You get a number, maybe a letter grade, and a vague list of suggestions. You have no idea what actually matters or why your score went up or down.
We think that defeats the purpose. So here is every factor Adymus checks, the exact weight each one carries, and the thresholds that determine your score. This methodology is based on published research: the Princeton GEO study (showing citations boost AI visibility by up to 40%), Ahrefs' analysis of 17M AI citations, and SE Ranking's study of 129K domains.
1. Entity Clarity (19%)
Why it matters: AI models need to understand who you are before they can recommend you. Pages with clear entity definitions are 2.3x more likely to be cited (SE Ranking, 129K domains).
What Adymus checks:
- Value proposition in the first 300 words
- Business name and services clearly stated
- Target audience identified
- H1 heading present and descriptive
- Meta description with entity information
- About statement or "who we are" section
Key thresholds: Missing value proposition in first 300 words = -20 points. No H1 heading = -15 points. No target audience = -10 points. Each missing signal compounds. A page with no entity signals scores 0.
2. Schema Markup (17%)
Why it matters: Structured data makes your content machine-readable. Pages with schema markup have 40% higher citation rates in AI Overviews.
What Adymus checks:
- JSON-LD structured data present
- Schema types used (Organization, Article, FAQPage, HowTo, Product, etc.)
- Organization schema completeness (name, description, logo, contact)
- FAQPage schema validation
- Schema nesting and relationship depth
Key thresholds: No schema at all = score 0. Single basic schema type = 30-40. Organization schema with full details = 60+. Multiple valid schema types with FAQPage = 80+. Malformed JSON-LD is treated the same as no schema.
3. Content Structure (14%)
Why it matters: AI models parse content hierarchically. Well-structured content with clear headings, logical flow, and readable prose is easier for models to extract and cite.
What Adymus checks:
- Heading hierarchy (H1 through H6, proper nesting)
- Word count and content depth
- Flesch Reading Ease score (target: 60-70, "Standard")
- AI-generated content detection
- Paragraph length and scannability
Key thresholds: Under 150 words = -30 points. Readability below 50 (hard to read) = -15 points. Skipped heading levels (H1 to H3, no H2) = -10 points. No subheadings at all on a long page = -20 points.
4. Trust Signals (10%)
Why it matters: AI models prioritize trustworthy sources. Trust signals tell models your content is credible and your business is real.
What Adymus checks:
- HTTPS (SSL certificate)
- Contact information (email, phone, address)
- Author credentials with tiered scoring (PhD/Dr > CEO/Founder > 5+ years experience)
- Privacy policy and terms of service
- Cross-channel presence (social profiles linked from your site)
Key thresholds: Missing HTTPS = -30 points. No contact information = -20 points. Author with PhD or MD credentials gets maximum trust score. No author attribution = -10 points.
5. Citation Worthiness (10%)
Why it matters: AI cites content it can quote directly. Pages with original statistics are 4x more likely to be cited. Data tables make content 4.1x more citable.
What Adymus checks:
- Statistics and numerical data points
- Original research or proprietary data
- Data tables with structured information
- Passage citability (self-contained, quotable paragraphs of 40-60 words)
- Source attribution for claims
Key thresholds: No statistics or data = score under 30. Three or more statistics with source attribution = 60+. Original data or research = 80+. Data tables with 3+ rows add significant bonus.
6. Content Freshness (8%)
Why it matters: 76.4% of most-cited pages updated within 30 days (Ahrefs, 17M AI citations). AI models heavily favor recent content.
What Adymus checks:
- Date parsing from multiple sources: meta tags, HTTP headers, visible dates, schema dateModified
article:modified_timeandLast-Modifiedheaders- Visible "Last Updated" dates on the page
Key thresholds: Updated 0-30 days ago = score 100. 31-90 days = 70. 91-180 days = 40. Over 180 days = 0. No detectable date = score 50 (uncertain).
7. FAQ Content (7%)
Why it matters: FAQPage schema has a 41% citation rate vs 15% without it (Relixir study via Frase.io). FAQ sections give AI ready-made answers to cite directly.
What Adymus checks:
- FAQ sections detected on the page (question-answer patterns)
- FAQPage schema markup present and valid
- Number and quality of Q&A pairs
- Answer completeness (self-contained, fact-rich answers)
Key thresholds: FAQ content without FAQPage schema = -30 points (missed opportunity). No FAQ content at all = score 0. FAQ section with valid schema = 70+. Five or more well-structured Q&A pairs with schema = 90+.
8. AI Crawler Access (5%)
Why it matters: 23% of websites accidentally block AI crawlers (Ahrefs). If bots cannot access your content, AI cannot cite you.
What Adymus checks:
- robots.txt rules for 13 AI bots: GPTBot, ChatGPT-User, Google-Extended, PerplexityBot, ClaudeBot, anthropic-ai, Bytespider, cohere-ai, Diffbot, FacebookBot, OAI-SearchBot, YouBot, and CCBot
- llms.txt file presence (emerging standard for AI-specific crawling instructions)
- Per-bot allow/block status
Key thresholds: All bots blocked = score 0. Each blocked bot reduces your score proportionally. All bots allowed = score 80. All bots allowed + llms.txt present = 100.
9. External Validation (3%)
Why it matters: Businesses in the top 25% for web mentions are 10x more likely to be cited by AI. External validation tells models you are a recognized authority.
What Adymus checks:
- Awards and recognition mentions on the page
- Wikipedia references (double weight because of AI training data importance)
- Media mentions and press coverage
- Brand citations from third-party sources
- Customer testimonials with named sources
Key thresholds: No external validation signals = score 30. Awards or media mentions = 60+. Wikipedia reference = significant bonus due to double weighting. Multiple validation signals from different categories = 80+.
10. Image Optimization (3%)
Why it matters: AI models use alt text and image context to understand page content. Well-described images add semantic meaning to your page.
What Adymus checks:
- Alt text quality (descriptive, 5-125 characters, not stuffed with keywords)
- Responsive images (srcset, sizes attributes)
- Modern image formats (WebP, AVIF)
- Image count relative to content length
Key thresholds: No images at all = score 80 (text-only pages are fine). Images without alt text = -15 points per image. Alt text under 5 characters or over 125 characters = flagged. All images optimized with descriptive alt text and modern formats = 90+.
11. Page Speed (2%)
Why it matters: Fast pages get 3x more citations than slow ones. AI crawlers have time budgets. If your page takes too long to load, it may not get fully indexed.
What Adymus checks:
- Time to First Byte (TTFB)
- Total content size
- Cache headers (Cache-Control, ETag)
- Mobile viewport meta tag
Key thresholds: TTFB over 800ms = critical. TTFB under 200ms = excellent. No cache headers = -15 points. Missing mobile viewport = -10 points. Content size over 5MB = flagged.
12. Sitemap Validation (2%)
Why it matters: A valid sitemap helps AI crawlers discover all your pages efficiently. It also signals which pages you consider important and when they were last updated.
What Adymus checks:
- XML sitemap present and well-formed
- lastmod dates on URLs (signals freshness to crawlers)
- Sitemap referenced in robots.txt
- URL count and structure
Key thresholds: No sitemap found = score 65. Sitemap present but no lastmod dates = 70. Valid sitemap with lastmod + robots.txt reference = 90+. Malformed XML = score 40.
How Your Overall Score Works
All 12 analyzers run in parallel when you scan a URL. Each produces a score from 0 to 100. These individual scores are multiplied by their weights and summed to produce your overall AI visibility score.
For example: if your Entity Clarity scores 80 and Schema Markup scores 60, those contribute (80 x 0.19) + (60 x 0.17) = 15.2 + 10.2 = 25.4 points toward your overall score. The same calculation runs for all 12 factors.
Adymus also generates separate per-platform scores for 8 AI platforms: ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini, Grok, DeepSeek, and Copilot. Each platform uses a different weight formula tuned to what that platform actually prioritizes. Your overall score tells you general AI readiness. Your per-platform scores tell you where to focus for specific platforms.
Frequently Asked Questions
Why do you publish your methodology?
Most AI visibility tools treat their scoring as a black box. We think that defeats the purpose. If you cannot see how your score is calculated, you cannot act on it. Publishing our methodology lets you verify our logic, prioritize your own fixes, and hold us accountable when something is off.
How often do the weights change?
We review weights quarterly based on new citation research. When a major study shifts the evidence (for example, the Ahrefs 17M citations study that confirmed the importance of freshness), we adjust. Weight changes are documented and communicated to users.
What score should I aim for?
Most websites score between 30 and 50 on their first scan. A score above 70 means you are well-positioned for AI citations. Above 85 puts you in the top tier. Focus on the highest-weighted factors first: Entity Clarity, Schema Markup, and Content Structure account for 50% of your total score.
Can I get a perfect 100?
It is possible but rare. A perfect score requires strong performance across all 12 factors, including signals like Wikipedia mentions and author credentials that take time to build. Most highly optimized sites score between 80 and 95.
Why does entity clarity have the highest weight?
AI models need to understand who you are before they can recommend you. If a model cannot identify your business, services, or audience, none of the other factors matter. Research from SE Ranking (129K domains) shows pages with clear entity definitions are 2.3x more likely to be cited.
How are per-platform scores different from the overall score?
The overall score uses the 12 weights listed in this article. Per-platform scores (ChatGPT, Perplexity, Google AI, Claude, Gemini, Grok, DeepSeek, Copilot) use different weight formulas tuned to what each platform actually prioritizes. For example, ChatGPT weights schema markup at 35%, while Perplexity prioritizes content depth and citations at 35% each.
Conclusion
Transparency builds trust. We publish our methodology because we believe you should know exactly how your score is calculated. If you disagree with a weight or threshold, we want to hear about it. That feedback makes the system better for everyone.
Run an Adymus scan to see where your site stands across all 12 factors. You will get a score for each factor, an overall AI visibility score, per-platform breakdowns for 8 AI search engines, and prioritized recommendations for what to fix first.
See Your Score Across All 12 Factors
Get your AI visibility score with a breakdown by factor. See exactly where you are strong, where you are losing points, and what to fix first.
Analyze my site