How BrandCited measures AI visibility. The approach, versioned.
BrandCited measures two things. First, how 9 AI engines cite, mention, and represent your brand. Second, whether your site makes it easy for those engines to do so. Every scan runs both layers.
The output is a 0-to-100 visibility score, a prioritized fix list, and per-engine citation data. Every number in your report traces back to a real observation from a real engine response or a real HTTP request against your site.
Two layers, run in parallel.
Layer 1
Citation testing
We send real user-intent queries to 9 AI engines and parse every response for brand mentions, citation type, position, sentiment, and competitor appearances.
Layer 2
Site audit
We crawl your site the way AI bots do and score it against our AI-specific rubric covering crawler access, structured data, content, trust, technical foundation, and entity recognition.
The site-audit layer reads the raw server HTML your pages return, the same bytes an AI crawler like GPTBot, ClaudeBot, or PerplexityBot receives. AI crawlers do not run JavaScript. Anything that only appears after a browser executes your scripts, schema injected late, content rendered client-side, is invisible to them, so we score your site the way they actually see it, not the way it looks in a browser.
This is why our audit catches gaps that lighter tools miss. A checker that loads your page in a headless browser sees content the AI never will, and a shallow fetch can miss structured data your server renders correctly. We read exactly what the engines read, then grade that.
The site-audit layer scores your site across 8 weighted categories, 92 individual checks in all. Each category contributes a fixed share of the 0-to-100 score. The weights below are the live values the scanner runs against.
| Category | Weight | What it measures |
|---|---|---|
| Crawler Access | 15% | Whether AI bots can find, read, and understand your site. robots.txt rules for GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and other AI crawlers, plus llms.txt and llms-full.txt signal files. |
| Structured Data | 15% | The JSON-LD schema markup AI engines use to extract facts about your brand. Validated against Schema.org spec, not just detected. |
| Content Quality | 20% | Depth, structure, clarity, and freshness of your content. Includes first-hand experience markers and citation of external authoritative sources. |
| Trust & Authority | 15% | Signals AI engines treat as trust indicators: reviews, about/team/contact pages, case studies, social profiles, press coverage, backlink profile. |
| Technical | 10% | Core Web Vitals field data, HTTPS, canonicals, Open Graph, sitemap quality, brand name consistency across surfaces. |
| Entity Presence | 10% | Whether AI engines recognize your brand as a distinct entity. Wikidata, Wikipedia, Google Knowledge Graph, G2/Capterra/Trustpilot coverage, and cross-surface name consistency. |
| Answer Prominence | 10% | How prominently your brand appears within AI responses — first-mention rate, citation link presence, share of voice versus competitors, AI Overview ownership. |
| Content Authenticity | 5% | E-E-A-T signals that distinguish first-hand experience from AI-generated boilerplate. Author identity, originality, source attribution density, editorial policy. |
Every engine we test, chosen to cover the mainstream AI assistant and AI search surfaces of 2026.
ChatGPT
OpenAI
Claude
Anthropic
Gemini
Perplexity
Perplexity
Grok
xAI
DeepSeek
DeepSeek
Llama
Meta
Google AI Overviews
Microsoft Copilot
Microsoft
The methodology is versioned with semantic versioning. Every scan record stores the version it ran under. Two scans are directly comparable only when both stamps match. When a breaking change ships, trend charts render a marker on the changeover date so score shifts are never mistaken for real movement on your site.
Current: v2.6.0 "Composite"
2026-06-14
Adds the multi-modal content richness check — the one on-page citability signal the rubric lacked versus public GEO research. Surfaced by benchmarking BrandCited against the open-source claude-seo skill: its GEO model weights multi-modal content (text + media) heavily, and our rubric had only alt-text coverage, an accessibility signal, not a richness one.
View changelog →Different platforms measure different things. BrandCited tests citation behavior across 9 AI engines plus our site audit. A tool that scores SEO alone, or monitors only ChatGPT, will produce different numbers. Our scores are not directly comparable to them.
Three possibilities. Your site changed. An AI engine's behavior drifted. Or the methodology itself updated. Every scan is stamped with the methodology version it ran under, and version changes appear as markers on your trend charts.
The scoring combines site-level signals with real engine responses you don't control. Gaming one surface rarely moves the overall score. Fixing real citation gaps does.
We monitor the engine fleet against a fixed canary query set and track response stability over time. Small drift is noted in the changelog. Larger drift triggers a version update with clear customer communication.
Patches ship as needed. Minor updates ship roughly monthly. Major updates ship at most quarterly, with advance notice for score-moving changes.
Ready to run your first audit?
Every scan is stamped with the current methodology version so you can compare results over time.
Start free scan