System Architecture

Section Overview
This section explains the conceptual architecture of AI LocalRank — how North Star 4.1 provides the evaluative framework and how the DxExA model operationalizes it.

North Star 4.1 Overview

The Unifying Vision

North Star 4.1 is the conceptual backbone of AI LocalRank. It represents our unified model for understanding how AI platforms evaluate and represent businesses.

AI platforms ask three questions about your business:

  1. Can I DISCOVER this business? — Does this entity exist in my knowledge?
  2. Do I have EVIDENCE to answer questions? — What information can I access and trust?
  3. Can I provide ACTIONABLE help? — Can I help the user do something?

The Three Dimensions

North Star 4.1 organizes AI evaluation into three core dimensions:

1. Discoverability (D)

"Can AI find this business?"

Discoverability measures whether AI platforms recognize your business as a distinct, real-world entity.

Key factors:

  • Does the business have a unique digital identity?
  • Is the business listed in authoritative directories?
  • Do multiple independent sources confirm the business exists?
  • Is there a knowledge graph entry (Wikipedia, Wikidata)?
  • Is NAP (Name, Address, Phone) consistent across the web?

What high Discoverability means: AI can identify your business confidently as a real, distinct entity — not confused with similar businesses, not fragmented across inconsistent listings.

What low Discoverability means: AI is uncertain whether your business exists, which business you are, or whether different mentions refer to the same entity.

2. Evidence (E)

"Can AI answer questions about this business?"

Evidence measures whether AI platforms have access to the information needed to answer user questions.

Key factors:

  • Is the website accessible and crawlable?
  • Is structured data (Schema.org) present and complete?
  • Is Google Business Profile claimed and populated?
  • Are reviews present on major platforms?
  • Is information fresh and current?

What high Evidence means: AI has the raw material to formulate accurate, confident answers about hours, services, contact information, and more.

What low Evidence means: AI lacks information, encounters gaps, or finds only partial data — leading to hedged answers, omissions, or errors.

3. Actionability (A)

"Can AI help users take action?"

Actionability measures whether AI platforms can guide users to the next step — calling, booking, navigating, or engaging.

Key factors:

  • Is a phone number clearly available?
  • Are booking or reservation links present?
  • Can AI provide directions (maps integration)?
  • Are email or contact options accessible?
  • Is the call-to-action path clear and functional?

What high Actionability means: AI can close the loop by connecting users with your business — "Here's their number" or "You can book here."

What low Actionability means: AI can describe your business but cannot help users take action — leaving them to search further on their own.

Why Three Dimensions?

North Star 4.1 uses three dimensions because AI confidence requires all three.

If ANY dimension fails, overall confidence collapses:

  • High D, High E, Low A = "I know about them but can't help you contact them"
  • High D, Low E, High A = "They exist but I don't know their hours/services"
  • Low D, High E, High A = "I'm not sure which business you mean"

ALL THREE must be strong for confident AI responses.

The DxExA Framework

From Concept to Execution

DxExA is the operational framework that implements North Star 4.1. The "x" represents multiplication — a deliberate choice that reflects how AI confidence works.

Final Confidence = D × E × A

Discoverability × Evidence × Actionability

Why Multiplication?

The multiplicative model reflects a fundamental truth about AI confidence:

If any factor is zero, overall confidence is zero.

Scenario D E A Result
Perfect on all dimensions 0.9 0.9 0.9 0.73 (Strong)
Missing Discoverability 0.0 0.9 0.9 0.00 (Failure)
Missing Evidence 0.9 0.0 0.9 0.00 (Failure)
Missing Actionability 0.9 0.9 0.0 0.00 (Failure)
Weak on one dimension 0.9 0.3 0.9 0.24 (Low)

This matches real AI behavior. A business that exists in directories but has no website content will fail Evidence checks. A business with great content but missing contact information will fail Actionability checks. AI platforms need all three to respond confidently.

Components of Each Dimension

Each DxExA dimension comprises multiple signals:

Discoverability (D) Components

  • Entity Strength: Does the business have a unique identity? Is there a Knowledge Graph presence? Is NAP consistent?
  • Corroboration: How many independent sources confirm the business? Is there cross-domain citation diversity? Do sources agree or conflict?
  • Authority: How established is the online presence? What is the domain authority level? Are there backlinks from reputable sources?

Evidence (E) Components

  • Site Quality: Is the website accessible? Is content substantive? Are there technical barriers?
  • Schema Quality: Is structured data present? How complete is the markup? Are required fields populated?
  • GBP Quality: Is Google Business Profile claimed? Are hours and services complete? Are photos present?
  • Reviews Quality: Are there reviews on major platforms? What is the volume and recency? What is the sentiment?
  • Recency: How fresh is the information? When was the data last updated?

Actionability (A) Components

  • Presence: Is a phone number available? Is a booking/reservation link present? Are maps/directions accessible? Is email contact available?
  • Convenience: Are action paths clearly structured? Can AI extract and present them easily? Is freshness of action data verified?

Platform-Specific Weights

Different AI platforms weight signals differently:

ChatGPT

  • High weight: Website content, Schema markup
  • Moderate weight: Google Business Profile
  • Lower weight: Reviews, Social signals

Perplexity

  • High weight: External citations, Multi-source agreement
  • Moderate weight: Reviews, Directory presence
  • Lower weight: Own website (requires corroboration)

Gemini

  • High weight: Google Business Profile, Schema.org
  • Moderate weight: Knowledge Graph, Reviews
  • Lower weight: Non-Google sources

Claude

  • High weight: Website quality, Consistency
  • Moderate weight: Wikipedia/Wikidata, Directories
  • Behavioral: Hedges when sources conflict

Grok

  • High weight: X (Twitter) presence, Social engagement
  • Moderate weight: Knowledge Graph, Real-time signals
  • Lower weight: Traditional directories

AI LocalRank applies these platform-specific weights through what we call "lenses" — platform behavior models that simulate how each AI would evaluate your business.

How AV Executes the Model

AV: The Execution Engine

AV (AI Visibility) is the module that executes North Star 4.1 and the DxExA framework. It is the deterministic engine that transforms collected evidence into platform-specific confidence scores.

AV Execution Flow:

  1. EVIDENCE COLLECTION — Gather all available business signals
  2. NORMALIZATION — Standardize data formats and resolve basics
  3. SCENARIO GENERATION — Create questions users might ask
  4. PLATFORM PROJECTION — Apply platform lenses to evaluate each scenario, compute D × E × A per platform per scenario
  5. AGGREGATION — Calculate rollup scores, detect conflicts, flags
  6. SNAPSHOT PUBLICATION — Freeze a complete diagnostic report

Key Characteristics of AV

Characteristic Description
Deterministic Same inputs always produce the same outputs
Evidence-based All scores derive from observable data
Platform-aware Each AI platform is modeled separately
Scenario-specific Evaluates specific questions, not abstract "visibility"
Diagnostic Produces explanations, not just scores

What AV Does Not Do

  • Does not call AI platforms directly — It simulates based on evidence
  • Does not modify your data — It observes and analyzes
  • Does not guess — It computes from evidence
  • Does not promise outcomes — It diagnoses current state

High-Level Data Flow

Complete System Diagram

The system flows from Business Reality through Evidence Collection to North Star 4.1 Evaluation:

BUSINESS REALITY

  • Website, Google Business Profile, Directories
  • Schema.org, Social Media, Reviews
  • Knowledge Graph, Listings, Citations

↓

EVIDENCE COLLECTION

  • Google Places API (GBP data, hours, reviews)
  • Website HTML Parser (Schema.org, contact info)
  • Directory Discovery (Yelp, TripAdvisor, etc.)
  • Knowledge Graph APIs (Wikidata, Wikipedia)
  • Citation Analysis (external mentions)

↓

NORTH STAR 4.1 EVALUATION

  • DxExA Framework (D: Discoverability, E: Evidence, A: Actionability)
  • Platform Lenses (ChatGPT, Perplexity, Gemini, Claude, Grok)

↓

AV MODULE (Execution Engine) + SUPPORTING MODULES

  • AV: Platform Scores, Scenario Results, Conflict Detection, Diagnostic Flags
  • Entity (Identity), Intent (User Goals), Access (Technical), Listings (Presence), Schema (Structure), AI Policy (Governance)

↓

AUDIT REPORT

  • Platform confidence scores
  • Scenario-by-scenario results
  • Diagnostic explanations
  • Conflict and gap detection
  • Power Plan recommendations

How This Fits in the System

The System Architecture is the foundation upon which all modules operate:

Module How It Relates to Architecture
AV Direct execution of DxExA framework
Entity Feeds Discoverability (D) dimension
Intent Interprets WHY AI makes choices given D×E×A
Access Gates whether Evidence (E) is even accessible
Listings Contributes to Corroboration within D
Schema Primary driver of Evidence (E) quality
AI Policy Governance layer for AI communication

What You Will See in the UI

The architecture manifests in your report as:

  • Platform Cards — Per-platform confidence scores derived from DxExA
  • Dimension Breakdowns — How D, E, and A contribute to each platform score
  • Scenario Results — How specific questions would be answered
  • Cross-Platform Comparison — Why platforms differ in confidence
  • Evidence Attribution — What data drives each score

Common Misunderstandings

Misunderstanding Clarification
"One high score is enough" All three dimensions must be strong
"It's just about having information" Structure and consistency matter as much as presence
"All platforms work the same" Each platform has distinct priorities and behaviors
"More is always better" Conflicting information is worse than missing information
"D×E×A is a simple formula" Each dimension has multiple weighted components

Clear Boundaries

This architecture section intentionally does not expose:

  • Exact numerical weights for each signal
  • Specific threshold values for scoring bands
  • Internal penalty calculation formulas
  • Platform lens implementation details

These are abstracted because they:

  1. Evolve as we refine the system
  2. Vary by business vertical
  3. Are calibrated through ongoing research

What we do expose:

  • The conceptual structure (North Star 4.1)
  • The evaluation dimensions (D×E×A)
  • The signal categories within each dimension
  • The platform differentiation model