SurfSense/.cursor/skills/entity-optimizer/references/knowledge-graph-guide.md

Knowledge Graph Optimization Guide

Part of entity-optimizer. See also: entity-signal-checklist.md

Comprehensive playbook for establishing and maintaining entity presence across Google Knowledge Graph, Wikidata, Wikipedia, and other knowledge bases.

How Knowledge Graphs Work

The Entity Web

Knowledge graphs are interconnected databases of entities and their relationships. Search engines and AI systems use them as ground truth for entity understanding.

Your Entity
├── is described by → Wikidata entry
├── is described by → Wikipedia article
├── is described by → Schema.org markup on your site
├── is linked to → Social profiles (LinkedIn, X, etc.)
├── is mentioned by → News articles, industry sites
├── is associated with → Topics, industries, other entities
└── is recognized by → Google Knowledge Graph, Bing Satori, AI training data

Which Knowledge Graphs Matter

Knowledge Graph	Who Uses It	Impact
Google Knowledge Graph	Google Search, Google AI	Powers Knowledge Panels, rich results, entity understanding in search
Wikidata	Google, Bing, Apple, Amazon, AI systems	Open data feeds multiple knowledge graphs; primary structured data source
Wikipedia	Google, all AI systems	Training data for every major LLM; Knowledge Panel descriptions often sourced here
Bing Satori	Bing, Copilot	Powers Bing's entity understanding and Microsoft Copilot
Schema.org (your site)	All search engines, AI crawlers	First-party structured data you control directly
DBpedia	Research, some AI systems	Auto-extracted from Wikipedia; relevant for academic/research entities

Data Flow

Your Website (Schema.org) ─┐
Wikidata ──────────────────┤
Wikipedia ─────────────────┼──→ Google Knowledge Graph ──→ Knowledge Panel
Industry Directories ──────┤                              AI Search Results
News/Media Mentions ───────┤                              Rich Results
Social Profiles ───────────┘

Understanding this flow is key: you influence the Knowledge Graph by controlling the source signals that feed it.

Google Knowledge Graph

Getting Into the Knowledge Graph

There is no "submit to Knowledge Graph" form. Google builds its Knowledge Graph from multiple sources. To get included:

1. Have a Wikidata entry — This is the most direct path
2. Earn a Wikipedia article — Strongest single signal
3. Implement Schema.org markup — Provides structured self-description
4. Get mentioned on authoritative sites — Third-party validation
5. Build branded search demand — Signals that users look for your entity

Checking Your Knowledge Graph Status

Method 1: Google Search Search for your entity name in quotes. If a Knowledge Panel appears on the right, you're in the Knowledge Graph.

Method 2: Knowledge Graph API

GET https://kgsearch.googleapis.com/v1/entities:search?query=[entity]&key=[API_KEY]

Response includes:

• @id: Your Knowledge Graph ID (e.g., kg:/m/0wrt4g)
• name: Entity name as Google understands it
• description: Short entity description
• detailedDescription: Longer description (usually from Wikipedia)
• resultScore: Confidence score (higher = more established entity)

Method 3: ~~knowledge graph If connected, query directly for entity status and attributes.

Claiming Your Knowledge Panel

1. Search for your entity on Google
2. If Knowledge Panel appears, look for "Claim this knowledge panel" link at bottom
3. Verify via official website, Search Console, YouTube, or other Google property
4. Once claimed, you can suggest edits (but Google has final say)

Common Knowledge Panel Fixes

Problem	Solution
No Knowledge Panel	Build Wikidata entry + Schema.org + authoritative mentions. Timeline: 2-6 months.
Wrong image	Update preferred image on: Wikidata (P18), About page, social profiles. Claim panel and suggest preferred image.
Wrong description	Edit Wikidata description. Update first paragraph of About page and Wikipedia article.
Missing attributes	Add properties to Wikidata and Schema.org. Claim panel and suggest additions.
Outdated information	Update Wikidata, About page, Wikipedia, and social profiles. Request refresh via claimed panel.
Wrong entity shown	Disambiguation needed. See Wikidata section below for disambiguation strategy.

Wikidata

Why Wikidata Is Critical

Wikidata is the single most influential editable knowledge base for entity optimization:

• Google uses it as a primary source for Knowledge Panels
• Bing uses it for Satori knowledge graph
• AI systems reference it during entity resolution
• It's open and you can edit it (within their guidelines)

Creating a Wikidata Entry

Step 1: Check Eligibility

Wikidata requires "notability" — the entity must be referenced in at least one external source. Unlike Wikipedia, the notability bar is lower: a company mentioned in a news article, a product with reviews, or a person with published work typically qualifies.

Step 2: Create the Item

1. Go to https://www.wikidata.org/wiki/Special:NewItem
2. Fill in:
   • Label: Official entity name
   • Description: Short description (e.g., "American software company" or "SEO optimization tool")
   • Aliases: Alternative names, abbreviations, former names

Step 3: Add Core Statements

Essential properties for each entity type:

Organizations:

Property	Code	Example
instance of	P31	business (Q4830453) or specific type
official website	P856	https://example.com
inception	P571	2020-01-15
country	P17	United States (Q30)
headquarters location	P159	San Francisco (Q62)
industry	P452	software industry (Q638608)
founded by	P112	[founder's Wikidata item]
CEO	P169	[CEO's Wikidata item]

Persons:

Property	Code	Example
instance of	P31	human (Q5)
occupation	P106	software engineer (Q183888)
employer	P108	[company Wikidata item]
educated at	P69	[university Wikidata item]
country of citizenship	P27	[country item]
official website	P856	https://example.com

Products/Software:

Property	Code	Example
instance of	P31	software (Q7397) or web application (Q189210)
developer	P178	[company Wikidata item]
official website	P856	https://example.com
programming language	P277	Python (Q28865)
operating system	P306	Linux (Q388)
software license	P275	Apache-2.0 (Q13785927)
inception	P571	2023-06-01

Step 4: Add External Identifiers

These link your Wikidata item to other knowledge bases:

Identifier	Code	Purpose
official website	P856	Primary web presence
X (Twitter) username	P2002	Social presence
LinkedIn organization ID	P4264	Professional presence
GitHub username	P2037	Technical presence
CrunchBase ID	P2087	Business data
Google Knowledge Graph ID	P2671	Google entity link
App Store ID	P3861	Mobile presence

Step 5: Add References

Every statement must have a reference. Unreferenced statements may be removed.

Good reference sources:

• Official website (for factual claims like founding date)
• News articles (for events, milestones)
• Industry reports (for market position)
• Government registries (for legal entity information)

Wikidata Maintenance

Task	Frequency	Why
Review existing statements	Quarterly	Ensure accuracy; update changed information
Add new properties	When new information available	Keep entry comprehensive
Check for vandalism	Monthly	Others can edit your entry
Add new references	When new coverage appears	Strengthen statement credibility
Update identifiers	When new profiles created	Keep links current

Wikipedia

Notability Requirements

Wikipedia requires entities to meet "general notability guidelines" (GNG):

• Significant coverage in reliable, independent sources
• Coverage must be non-trivial (not just a mention or directory listing)
• Sources must be independent of the entity (not press releases, not entity's own content)

Building Toward Notability

If the entity doesn't have a Wikipedia article yet:

1. Audit existing coverage: Search Google News, academic databases, and industry publications for mentions
2. Identify gaps: What kinds of coverage are missing?
3. Build coverage first, then article: The article is the last step, not the first

Coverage-building strategies:

Strategy	Timeline	Notability Impact
Industry report mentions	3-6 months	Medium — depends on report authority
News article coverage	1-3 months	High — especially from recognized publications
Conference speaking + coverage	3-12 months	Medium — needs post-event coverage
Academic paper citations	6-12+ months	High — very strong for GNG
Award recognition	Variable	Medium — depends on award authority
Book publication or feature	6-12+ months	High — strong independent source

Wikipedia Article Best Practices

DO:

• Write in neutral, encyclopedic tone
• Use only independent, reliable sources as references
• Follow Wikipedia's Manual of Style
• Disclose any conflict of interest on your Talk page
• Let the community review and improve the article

DO NOT:

• Write promotional content
• Use the entity's own website as a primary source
• Create the article from a company account without disclosure
• Remove criticism or negative but sourced information
• Pay someone to write the article without disclosure (violates Wikipedia policy)

Wikipedia's Impact on AI

Wikipedia is disproportionately important for AI systems because:

• It's in the training data of every major LLM
• AI systems treat it as a high-trust source
• Wikipedia's structured format makes it easy for AI to extract and cite
• The first paragraph of a Wikipedia article often becomes the AI's entity definition

This makes Wikipedia presence one of the highest-impact entity optimization actions for GEO.

Schema.org Entity Markup

Minimum Viable Entity Schema

Every entity should have at minimum this markup on the homepage:

Organization:

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://example.com/#organization",
  "name": "Example Corp",
  "url": "https://example.com",
  "logo": "https://example.com/logo.png",
  "description": "Example Corp is a [what it is] that [what it does].",
  "foundingDate": "2020-01-15",
  "founder": {
    "@type": "Person",
    "name": "Jane Smith",
    "@id": "https://example.com/about/jane-smith#person"
  },
  "sameAs": [
    "https://www.wikidata.org/wiki/Q12345678",
    "https://en.wikipedia.org/wiki/Example_Corp",
    "https://www.linkedin.com/company/example-corp",
    "https://x.com/examplecorp",
    "https://www.crunchbase.com/organization/example-corp"
  ]
}

Person:

{
  "@context": "https://schema.org",
  "@type": "Person",
  "@id": "https://example.com/about/jane-smith#person",
  "name": "Jane Smith",
  "url": "https://example.com/about/jane-smith",
  "image": "https://example.com/photos/jane-smith.jpg",
  "jobTitle": "CEO",
  "worksFor": {
    "@type": "Organization",
    "@id": "https://example.com/#organization"
  },
  "description": "Jane Smith is [who they are] specializing in [expertise areas].",
  "sameAs": [
    "https://www.wikidata.org/wiki/Q87654321",
    "https://www.linkedin.com/in/janesmith",
    "https://x.com/janesmith"
  ]
}

sameAs Best Practices

The sameAs property is the primary entity disambiguation signal in Schema.org. It tells search engines "this is the same entity as the one on these other platforms."

Must include (when available):

1. Wikidata URL (most important for Knowledge Graph)
2. Wikipedia URL
3. LinkedIn URL
4. Official social media profiles

Include when relevant: 5. CrunchBase URL 6. GitHub URL 7. IMDb URL (for people in entertainment) 8. Industry directory URLs

Common mistakes:

• Linking to generic pages instead of entity-specific URLs
• Inconsistent: Schema says "Example Corp" but LinkedIn says "Example Corporation"
• Missing Wikidata link (this is the single most impactful sameAs)
• Including dead or redirecting URLs

Cross-Page Entity Consistency

Every page on the site should reference the same entity with the same @id:

{
  "@type": "WebPage",
  "publisher": {
    "@type": "Organization",
    "@id": "https://example.com/#organization"
  }
}

For articles:

{
  "@type": "Article",
  "author": {
    "@type": "Person",
    "@id": "https://example.com/about/jane-smith#person"
  },
  "publisher": {
    "@type": "Organization",
    "@id": "https://example.com/#organization"
  }
}

This creates a consistent entity graph that search engines can confidently map to Knowledge Graph entries.

Monitoring Entity Health

Quarterly Entity Health Check

Check	How	What to Look For
Knowledge Panel accuracy	Google entity name	Correct info, image, attributes
Wikidata entry	Visit Wikidata page	No vandalism, info still current
AI entity resolution	Query 3+ AI systems	Accurate recognition and description
Schema.org validation	Google Rich Results Test	No errors, complete entity data
Branded search SERP	Google "[entity name]"	Clean SERP, no disambiguation issues
Social profile consistency	Visit all profiles	Same name, description, links

Entity Health Metrics to Track

Metric	Tool	Target
Knowledge Panel presence	Google Search	Present and accurate
Branded search CTR	~~search console	> 50% for exact brand name
AI recognition rate	Manual testing	Recognized by 3/3 major AI systems
Wikidata completeness	Wikidata	15+ properties with references
Schema.org error count	Google Search Console	0 errors
Brand mention volume	~~brand monitor	Stable or growing trend

Recovery Playbooks

Entity disappeared from Knowledge Graph:

1. Check if Wikidata entry was deleted or merged
2. Verify Schema.org markup hasn't changed
3. Look for major algorithm updates that might have affected entity recognition
4. Rebuild signals: start with Wikidata, then Schema.org, then external mentions
5. Timeline: 2-8 weeks for recovery

AI systems giving incorrect entity info:

1. Identify which sources have incorrect information
2. Correct information at source (Wikidata, Wikipedia, About page)
3. AI systems will update over time (training data refresh + live search)
4. For urgent issues, some AI systems have feedback mechanisms
5. Timeline: weeks to months depending on AI system update cycles

Knowledge Panel showing wrong entity:

1. Claim the Knowledge Panel (if you haven't already)
2. Strengthen disambiguation signals (see SKILL.md Disambiguation Strategy)
3. Add qualifier to entity name if needed
4. Build more unique entity signals (original content, specific topic associations)
5. Timeline: 1-3 months