MAVI Methodology Note v1.0: Definitions, Formula, Missing Data and Reproducibility

Author & Researcher: Taqi Molavi Publisher: Molavi R&D Think Tank (molavi.pro/notes) Publication Date: October 10, 2026 Version: 1.1 (Technical Methodology Note) Category: AI Visibility Framework & Applied Mathematics

1. Architectural Purpose and Design Philosophy

The Molavi AI Visibility Index (MAVI) is an operational diagnostic and measurement framework developed to replace arbitrary, unscientific "AI SEO Scores" with a mathematically disciplined standard. This technical note specifies the composite scoring formulation, empirical calibration of layer weights, handling of incomplete data sets, and the closed-loop remediation pipeline powered by our open-source tools.

2. Composite Formulation and Calibrated Baseline Weights

The composite MAVI score is computed as a normalized weighted linear sum across five operational layers:

$$\text{MAVI} = \sum_{i=1}^{5} w_i \times L_i \quad \text{subject to} \quad \sum_{i=1}^{5} w_i = 1.0$$

Calibrated baseline weights established through multi-corpus testing across 50 enterprise domains:

$$\text{MAVI} = 0.15(L_1) + 0.20(L_2) + 0.20(L_3) + 0.20(L_4) + 0.25(L_5)$$

Mathematical Rationale:

* $w_1 = 0.15$ (Layer 1: Technical Accessibility): Clean DOM and machine readability are foundational prerequisites; failure here prevents indexing, but technical perfection alone confers no semantic advantage. * $w_2 = 0.20$ (Layer 2: Semantic Extractability): Structured Schema.org graph markup enabling frictionless token ingestion. * $w_3 = 0.20$ (Layer 3: Entity Clarity & Chunk Fitness): Vector embedding cohesion and knowledge graph disambiguation. * $w_4 = 0.20$ (Layer 4: Citation Readiness / CSP): Citation survival proxy, independent verification, and standalone proposition density. * $w_5 = 0.25$ (Layer 5: Empirical AI Visibility): Highest weighting allocated to observed, multi-run performance in live frontier models under zero-prompt conditions.

3. Structural Readiness Decoupling ($\text{MAVI-R}$)

A foundational architectural requirement is the strict decoupling of code-level readiness from live model behavior:

$$\text{MAVI-R} = \frac{w_1(L_1) + w_2(L_2) + w_3(L_3) + w_4(L_4)}{w_1 + w_2 + w_3 + w_4} = \frac{0.15(L_1) + 0.20(L_2) + 0.20(L_3) + 0.20(L_4)}{0.75}$$

This enables engineering organizations to benchmark deployment readiness deterministically before executing stochastic live model evaluations.

4. Missing Data Policy

When live-model benchmarking is unavailable due to budget, latency, or API access constraints:

1. Zero-Defaulting Prohibited: Layer 5 is never assigned a score of 0. In statistical measurement, a zero indicates a failed test, not an unmeasured condition.
2. Explicit Uncertainty Labeling: The layer is designated as not_measured.
3. Isolated Readiness Reporting: Under partial data, only the Readiness Score ($\text{MAVI-R}$) is published, preventing the creation of distorted composite figures.

5. Worked Numerical Case Study

Consider an enterprise SaaS provider audited with SAGE and GEO-Scope: * $L_1 = 90$ (Flawless SSR, fast response, pristine headers) * $L_2 = 85$ (Comprehensive JSON-LD Organization and Product schemas) * $L_3 = 70$ (Verified Wikidata identity, but conflicting secondary directories) * $L_4 = 55$ (Low factual token density, promotional marketing copy) * $L_5 = 35$ (Suppressed citation rate across 100 live benchmark runs)

Step 1: Compute Structural Readiness Score ($\text{MAVI-R}$): $$\text{MAVI-R} = \frac{0.15(90) + 0.20(85) + 0.20(70) + 0.20(55)}{0.75} = \frac{13.5 + 17.0 + 14.0 + 11.0}{0.75} = \frac{55.5}{0.75} = \mathbf{74.0\%}$$

Step 2: Compute Composite MAVI Score: $$\text{MAVI} = 55.5 + 0.25(35) = 55.5 + 8.75 = \mathbf{64.25\%}$$

Engineering Diagnosis: While technically robust (74.0% readiness), the platform underperforms in live models (35.0%) due to low proposition density ($L_4 = 55$). Remediation requires VPA passage restructuring rather than further technical optimization.

6. The Closed-Loop Open-Source Ecosystem

To ensure transparency and reproducibility without proprietary lock-in:

$$\text{AnswerPath} \longrightarrow \text{GEO-Scope} \longrightarrow \text{SAGE} \longrightarrow \text{MAVI} \longrightarrow \text{SiteProbe} \longrightarrow \text{Re-measurement}$$

1. [AnswerPath](https://github.com/tmolavi/answerpath-geo): Unbranded intent mining and query clustering.
2. [GEO-Scope](https://github.com/tmolavi/geo-scope): Stochastic multi-engine benchmarking and confidence intervals.
3. [SAGE](https://github.com/tmolavi/sage-audit): Automated code and schema audit engine for $L_1$–$L_3$.
4. [mcp-geo-server](https://github.com/tmolavi/mcp-geo-server): Model Context Protocol interface for coding agents.
5. [SiteProbe](https://github.com/tmolavi/siteprobe): Autonomous DOM and metadata remediation.

7. Primary References & Next Steps

* Comprehensive White Paper: The Molavi AI Visibility Index (MAVI) * Foundational Strategic Pyramid: The Molavi GEO Pyramid * Case Study Archive: Technical Proof & Evidence * Commission an Institutional Audit: Enterprise GEO & AI Visibility Audits