How It Works
How LLM Check measures AI Visibility
LLM Check runs live queries across ChatGPT, Gemini, and Perplexity to find out whether AI engines mention your brand — and analyses your page to tell you why they do or don't.
Two measurements, one picture
LLM Check produces two distinct scores that work together. The AI Visibility Rate (formerly known as the Weighted Visibility Score, or WVS) is the live measurement — what AI engines actually said when asked about your space. The Structural Markup Score is the diagnostic — a page-level audit that explains why your visibility is where it is and what to change.
AI Visibility Rate
% of queries where AI mentioned you
Structural Markup Score
0–100 diagnostic from page content
AI Visibility scan
Each scan generates 20 queries about your brand, tailored to your industry and content using Claude. Every query is sent to all three AI engines in parallel, producing 60 independent results per scan. A result counts as visible if the engine mentioned your brand name or cited your domain in its response.
20
queries per scan
×3
AI engines
60
data points
AI Visibility Rate
The rate is the percentage of the 60 data points where your brand appeared. A result counts as visible under either condition — the engine cited your domain by URL, or the engine mentioned your brand name in its response. Perplexity and ChatGPT share source URLs; Gemini responses are scored on mention only.
Denied responses are treated separately: when an engine explicitly says it cannot find your brand, that response is flagged as “Denied” rather than a plain miss. It is not counted as a visible result, but it surfaces as a distinct amber signal — an engine that can't find you is a different problem from one that simply didn't mention you.
A note on our history: when LLM Check launched, this measurement was a composite metric we called the Weighted Visibility Score (WVS) — a weighted blend of mention rate, prominence, sentiment alignment, and source provenance. We retired it because a single transparent percentage beats a weighted composite for trust: you can check our math by reading the scan results yourself.
Verdict bands
The rate maps to one of four verdicts. These bands are the same across every surface in LLM Check — the dashboard, the sites list, shared reports, and PDFs. A 65% site is At Risk everywhere, never green in one place and amber in another.
| Verdict | AI Visibility Rate | Meaning |
|---|---|---|
| Strong Visibility | ≥ 80% | AI engines reliably recommend you across query types |
| At Risk | 60 – 79% | Visible in most queries but gaps exist that competitors can exploit |
| Low Visibility | 40 – 59% | Appearing in fewer than half of relevant AI answers |
| Rarely Visible | < 40% | AI engines rarely surface your brand — significant opportunity |
4 query types, 5 queries each
The 20 queries are split across four intent categories that reflect how real users ask AI engines about your category. Results are broken down by type so you can see where you're strong and where you're being missed.
Brand
"What is [your brand]?" / "Tell me about [brand]"
Queries that name your brand directly. Low visibility here means AI engines don't know who you are.
Problem
"How do I [problem your product solves]?"
Queries describing the pain point. Missing here means AI engines answer the problem without recommending you.
Category
"Best [product category]" / "Top [tools] for [use case]"
Category-level queries. These drive the most traffic — visibility here is often the highest leverage.
Comparison
"[Your brand] vs [competitor]" / "[Brand] alternatives"
High-intent queries where buyers are close to a decision. Missing here cedes the final mile to competitors.
Structural Markup Score
The structural score (0–100) is computed by Claude, which reads your page content and evaluates six parameters that research links to AI citation likelihood. The score is a weighted average — higher-weight parameters matter more. It tells you what to fix, not just what your rate is.
| Parameter | Weight | What it measures |
|---|---|---|
| Answer-Ready Content | 25% | Whether your content directly answers questions AI users ask — structure, headings, FAQs, and concise answers |
| Brand & Expertise Clarity | 20% | How clearly the brand name, what it does, and who it serves are stated on the page |
| Structured Data Depth | 20% | Schema.org markup quality — FAQPage, Organization, BreadcrumbList, and other types AI engines use to extract facts |
| Citable Content Quality | 15% | Whether your content contains quotable claims, data points, or specific facts that AI engines can attribute to you |
| E-E-A-T & Authority Signals | 10% | Experience, Expertise, Authoritativeness, and Trust signals — author credentials, sourcing, and external recognition |
| Comparison & Intent Coverage | 10% | Whether the site addresses comparison queries, alternatives, and use-case breadth across the category |
Free accounts see the first three parameters. Premium unlocks all six and the full AI Visibility scan.
The three AI engines
Each engine has a different approach to sourcing and citation. LLM Check queries all three so your visibility rate reflects how you perform across the AI search landscape, not just one engine's perspective.
ChatGPT
GPT-4o with search (OpenAI)
Shares source URLs for cited content. Widely used for research and product discovery.
Gemini
gemini-2.5-flash (Google AI)
Scored on brand mention. Strong for category-level queries and brand awareness.
Perplexity
sonar model (Perplexity AI)
Shares detailed source URLs. The highest-signal engine for citation-source analysis.
See how your site performs
Run a free structural analysis to get your markup score, then unlock the AI Visibility scan to see how ChatGPT, Gemini, and Perplexity actually answer questions in your category.