On this page
01 · What the LVI measures
02 · Prompt corpus design
03 · Composite scoring
04 · Engine weights
05 · Score interpretation
06 · How to improve your score
07 · FAQ
What the LVI measures — and what it does not
The LLM Visibility Index measures one thing: how often and how prominently a brand is cited by AI engines in response to buyer-intent queries in its sector. It does not measure brand awareness, content quality, Google ranking, or social media presence. A brand can have excellent SEO and score 15/100 on the LVI. A brand with no website can score 60/100 if it is consistently cited in AI responses on relevant prompts.
| What LVI measures | What LVI does not measure |
|---|
| Citation frequency on buyer-intent prompts | General brand awareness |
| Citation position (1st vs 5th in response) | Google search ranking |
| Engine coverage (1/4 vs 4/4 engines) | Content quality or readability |
| Share of voice vs named competitors | Social media presence |
| Month-over-month progression | Website traffic or SEO metrics |
Prompt corpus design — how queries are selected
Each sector in the LVI has a corpus of 50+ buyer-intent prompts, refreshed monthly. Prompt selection criteria: (1) the query must represent a real purchasing decision, not an informational search — 'what is XDR' is excluded, 'which XDR for NIS2 French SME' is included, (2) the query must have at least 100 estimated monthly searches across AI engines combined, (3) the query must be specific enough to generate different answers for different brands — generic queries that always produce the same 3 global leaders are excluded, (4) the query corpus must cover all stages of the buying journey: category definition (30%), comparison (40%), vendor selection (30%).
Corpus refresh: 40-60% of Perplexity citations change month-over-month as it prioritizes real-time sources. The LVI prompt corpus is re-executed monthly to capture this volatility. A brand's score can shift by 5-15 points between editions based on distribution activity.
Composite scoring — how the 0-100 score is calculated
The composite LVI score weights three dimensions:
| Dimension | Weight | Definition | Why it matters |
|---|
| Citation frequency | 50% | % of corpus prompts where brand appears across all 4 engines | Core signal — appears vs does not appear |
| Citation position | 30% | Weighted average rank within response — 1st citation = 10pts, 5th = 2pts | Commercial impact differs by position |
| Engine coverage | 20% | Number of engines citing brand / 4 — 4/4 = full score | Breadth signals structural authority not tied to one engine |
Engine weights and per-engine scoring
All four engines are weighted equally in the composite score (25% each), but tracked separately. This matters because divergence between engines is an optimization signal. A brand at 80/100 on Perplexity and 20/100 on ChatGPT has a distribution problem, not a content problem. The LVI surfaces these gaps explicitly in the per-engine breakdown.
| Engine | Model | Update frequency | Primary citation signal |
|---|
| ChatGPT | GPT-4o | Corpus: 3-18 month lag | Historical DA, co-citations, institutional sources |
| Perplexity | Sonar Pro | Real-time + 30-day recency | Freshness, sector media, Reddit, LinkedIn |
| Gemini | Gemini 2.5 Pro | Corpus + Google index | Google ecosystem, Knowledge Graph, news |
| Claude | Haiku 4.5 | Corpus + MCP real-time | Editorial quality, institutional authority, MCP data |
How to interpret your LVI score
| Score range | Interpretation | Priority |
|---|
| 85-100 | Category default — cited in most responses on generic + specific prompts | Maintain and monitor competitors |
| 70-84 | Sector leader — consistent citation on mid-specificity prompts | Expand to long-tail and adjacent sectors |
| 50-69 | Competitive — cited on specific prompts, inconsistent on generic | Accelerate distribution velocity |
| 30-49 | Marginal — citation on niche prompts only, absent from selection queries | Corpus audit + tier-1 media distribution |
| 0-29 | Invisible — below citation threshold on buyer-intent prompts | Baseline audit before any content investment |
How to improve your LVI score — the three levers
Lever 1 — Distribution reach: publish answer-first content in sources LLMs cite (tier-1 sector media, Reddit, LinkedIn Pulse). This impacts Perplexity in 5-21 days and ChatGPT in 30-90 days. Lever 2 — Source diversity: citations from 5 different sources on the same prompt are worth more than 5 citations from 1 source. Reddit + media + LinkedIn + institutional = multi-source authority. Lever 3 — Engine-specific gaps: use LVI per-engine breakdown to identify which engine is underperforming. Allocate production to the source types that engine prioritizes.
Frequently asked questions
What is the LLM Visibility Index?
The LVI is a public benchmark measuring AI citation scores for 312 brands across 15 B2B sectors on ChatGPT, Perplexity, Gemini, and Claude. It runs a corpus of 50+ buyer-intent prompts per sector monthly and produces a composite score (0-100) based on citation frequency, position, and engine coverage.
How often is the LVI updated?
Monthly. The corpus is re-executed at the beginning of each month. Scores reflect the citation state for that month. Individual brands can see movements of 5-15 points between editions based on distribution activity.
Who builds and operates the LVI?
The LVI is built and operated by ZivRank, a B2B GEO agency specializing in regulated sectors (NIS2, DORA, CSRD). It is publicly accessible at llm-visibility-index.com. ZivRank uses the LVI as the primary measurement tool for all client engagements.
Is my brand automatically in the LVI?
No. Brands are added to the LVI when their citation frequency crosses a minimum threshold on sector-specific prompts. If your brand is not in the index, you can request a free Quick Scan at zivrank.com to receive a baseline score and gap analysis.
Can my LVI score decrease?
Yes. The LVI measures current citation patterns. If your competitors increase their distribution velocity, their scores increase and yours may decrease in relative Share of Voice terms. Absolute score decreases occur when Perplexity refreshes its RAG sources and drops content that is no longer recent enough.
Sources
- LLM Visibility Index — llm-visibility-index.com — July 2026
- ZivRank LVI methodology v2.1 — zivrank.com/methodology