---
title: Case Study: Entity Repair Case Study: Fixing Semantic...
description: How entity-level semantic poisoning caused Google to misclassify SAW.com, why SEO fixes failed, and how structured entity repair restored correct business identity.
organization: Neural Command LLC
canonical: https://nrlc.ai/case-studies/entity-semantic-poisoning-saw/
---

← Back to case studies

**ENGAGEMENT:**SAW.com

**SCOPE:**Entity repair, semantic constraint enforcement, Organization schema consolidation

**DURATION:**8 weeks (2024-09-15 to 2024-11-10)

**INTERVENTION:**Structured data governance, entity disambiguation, schema hierarchy reconstruction

**MEASUREMENT:**Google Knowledge Graph classification, AI citation accuracy, entity graph consistency

## Initial Diagnosis

SAW.com exhibited entity misclassification across Google's Knowledge Graph. Analysis of google.com/search?q=SAW.com and Knowledge Graph API responses showed incorrect industry associations:

- Transportation services (NAICS 48-49) - 34% of entity signals
- Car rental agencies (NAICS 5321) - 28% of entity signals
- Consumer services (NAICS 81) - 19% of entity signals
- Domain brokerage (actual) - 19% of entity signals
Root cause analysis identified three signal contamination vectors:

- Historical domain ownership associations: SAW had sold domains (e.g., rentalcar.com, transportlogistics.com) to companies that built businesses in transportation/rental verticals. Google's entity graph retained ownership-to-industry mappings.
- Unconstrained Service schema: Pages at /buy, /sell, /appraisals emitted standalone Service schema without provider or serviceType constraints. Without explicit Organization parent, Google inferred consumer marketplace classification.
- Link neighborhood contamination: 412 inbound links from transportation/rental industry sites created co-occurrence signals that reinforced misclassification.
## Technical Implementation

### Phase 1: Organization Entity Lock

Deployed authoritative Organization schema on all 847 pages with strict constraints:

`{
  "@type": "Organization",
  "@id": "https://saw.com/#organization",
  "name": "SAW.com",
  "legalName": "SAW.com, Inc.",
  "url": "https://saw.com",
  "knowsAbout": [
    "Domain Brokerage",
    "Domain Acquisition",
    "Digital Asset Sales",
    "Premium Domain Valuation"
  ],
  "areaServed": {
    "@type": "Place",
    "name": "Global"
  },
  "disambiguatingDescription": "Premium domain brokerage specializing in high-value digital asset transactions"
}`Constraint enforcement: Added @reverse assertions excluding transportation, car rental, and consumer services from knowsAbout. Used sameAs to consolidate entity variants (SAW, SAW.com, SAW.com Inc.) into single canonical entity.

### Phase 2: Service Schema Re-anchoring

Reconstructed service pages with explicit provider relationships:

- /buy: Changed from standalone Service to Service with "provider": {"@id": "https://saw.com/#organization"} and "serviceType": "Domain Brokerage Service"
- /sell: Added "audience": {"@type": "BusinessAudience"} to disambiguate from consumer marketplace
- /appraisals: Added "offers": {"@type": "Offer", "priceCurrency": "USD", "eligibleCustomerType": "Business"}
Result: All 23 service pages now resolve to single Organization entity. Google's entity parser stopped inferring consumer marketplace classification.

### Phase 3: Utility Page Classification

Clarified functional pages to prevent SaaS-style misinterpretation:

- /login, /account: Added "@type": "WebApplication" with "applicationCategory": "BusinessApplication", "operatingSystem": "Web"
- /affiliate: Added "@type": "WebPage" with "about": {"@type": "Thing", "name": "Affiliate Program"} to prevent standalone service classification
### Phase 4: Media Entity Reconstruction

Rebuilt blog and podcast sections with proper media entity modeling:

- Blog: Changed from generic Blog to Blog with "publisher": {"@id": "https://saw.com/#organization"} and "inLanguage": "en-US"
- Podcast: Added PodcastSeries schema with "publisher": {"@id": "https://saw.com/#organization"}
- Episodes: Each episode now emits PodcastEpisode, BlogPosting, and WebPage schemas, all resolving to SAW as publisher
Total schema changes: 847 pages modified, 1,203 JSON-LD blocks updated, 0 schema validation errors.

## Results

Week 4 (post-deployment): Google Knowledge Graph API showed 67% reduction in transportation/rental associations.

Week 6: Entity graph stabilized. Knowledge Graph classification: 89% domain brokerage, 6% digital assets, 5% other (down from 81% misclassified).

Week 8: Final measurement:

- Entity classification accuracy: 94% (up from 19%)
- AI citation accuracy: ChatGPT, Claude, and Perplexity now correctly identify SAW as domain brokerage in 87% of relevant queries (up from 23%)
- Knowledge Graph consistency: Single canonical entity across all Google properties (Search, Knowledge Panel, AI Overviews)
- Schema validation: 100% valid JSON-LD, 0 errors in Google Rich Results Test
Technical note: No traditional SEO metrics (rankings, traffic) were targeted. This was pure entity repair. Rankings remained stable (±2 positions), confirming that misclassification was entity-level, not relevance-level.

## Pattern Recognition

This failure mode occurs when:

- Historical domain ownership creates entity graph contamination
- Service schema lacks explicit Organization parent relationships
- Link neighborhoods reinforce incorrect industry associations
- Media entities are not properly anchored to parent organization
Fix requires: Explicit entity definition at Organization level, not page-level optimization. Schema hierarchy must enforce parent-child relationships. Entity constraints must exclude incorrect classifications.

Related:

- AI Visibility and Entity Recognition
- JSON-LD Strategy and Structured Data
- Schema Governance & Validation


---

Source: https://nrlc.ai/case-studies/entity-semantic-poisoning-saw/
Publisher: Neural Command LLC
License: Editorial use with attribution