MERCURY GEO METHODOLOGY

Three-Layer GEO Architecture

The only audit system that measures potential, validates with code, and proves results with actual AI citations.

// THE PROBLEM

The Problem with Single-Point Audits

Most GEO audits give you one number. That number cannot tell you if you are theoretically optimized but practically invisible.

Score Without Proof

You score 85/100. But are you actually cited by ChatGPT? No one checks.

Fixes Without Validation

Add FAQ schema. But did it work? Most audits never re-measure.

Theory vs Reality

A perfect technical score means nothing if AI platforms do not crawl you.

// THE ARCHITECTURE

Three Layers. One Truth.

We measure your AI visibility from three angles — code, content, and actual citations — then sync them into a single, validated score.

L3

Layer 3: Actual

45% Weight

Does AI actually cite us when users ask?

Method

  • Automated queries across ChatGPT, Perplexity, Gemini, Bing
  • 80 queries/day x 30 days = 2,400 observations/month
  • Captures citation position, excerpt, platform spread

Measures

  • Citation rate: % of queries where brand appears
  • Average position: 1st, 2nd, 3rd mention
  • Platform spread: cited on 4/4 platforms?
  • Excerpt quality: positive vs neutral mention
L2

Layer 2: Potential

35% Weight

Could AI cite us? Are we optimized?

Method

  • Full-site crawl (up to 200+ pages)
  • Schema validation and content analysis
  • 6-dimension diagnostic scoring

Measures

  • AI Citability (25%): FAQ, HowTo, structured content
  • Content Quality (20%): Depth, data, freshness
  • E-E-A-T (20%): Author bios, credentials, trust
  • Technical GEO (15%): Crawlers, speed, schema
  • Brand Authority (10%): Wikidata, sameAs
  • Platform Optimization (10%): llms.txt, multilingual
L1

Layer 1: Foundation

20% Weight

Is our code built to be AI-readable?

Method

  • Component-level code audit
  • Next.js SSR architecture review
  • Payload analysis per route

Measures

  • Content bundling (730KB to 175KB fix)
  • Schema implementation quality
  • JSON-LD validity and completeness
  • robots.txt, sitemap, llms.txt correctness
// THE SYNC

The Sync: How Layers Validate Each Other

// Week 1: Baseline
L2_Potential = 88   // Strong technical optimization
L3_Actual   = 67   // But AI does not cite us much
// → 21-point gap = discovery problem

// Action: Add FAQ schema to 39 service pages
await fixServicePages();

// Week 3 (after AI re-crawl cycle)
L2_Potential = 93   // +5 from fixes
L3_Actual   = 78   // +11 from improved citations
// → Causal validation: the fix worked

// True GEO Score
TrueScore = L1×0.20 + L2×0.35 + L3×0.45
          = 85      + 93      + 78
          = 84.15"Grade B+ (Strong)"
// THE FORMULA

The True GEO Score

Actual × 0.45 + Potential × 0.35 + Foundation × 0.20

Not all layers are equal. Actual citations (ground truth) carry the most weight. Potential (diagnostics) tells you what to fix. Foundation (code) is necessary but not sufficient.

20%
Foundation (Code Audit)
35%
Potential (GEO Audit v3)
45%
Actual (Citation Fan-Out)
// WHY IT WORKS

Why This Architecture Wins

Causal Validation

Unlike traditional audits, we prove that fixes work. L2 to L3 sync creates a feedback loop that improves accuracy over time.

Proprietary Data Moat

2,400 observations/month builds a dataset no competitor has. Trend data becomes your competitive advantage.

Actionable Diagnostics

L2 tells you exactly what is broken. L3 tells you if fixing it worked. No more guessing.

Client Lock-In

Once you have 6 months of L3 data, switching to a competitor means losing trend context and causal history.

Stop Guessing. Start Measuring.

Get the only audit system that connects technical optimization to actual AI citations — with proof.

Request a Three-Layer GEO Audit

Starting at $2,500/month · Results in 14 days · Cancel anytime