Martha van Berkel - From Schema Markup to Knowledge Graphs: Powering AI with Connected Data

Transcript

1. From Schema Markup to Knowledge Graphs Powering AI with Connected Data Martha van Berkel CEO, Schema App
2. Martha van Berkel’s Knowledge Graph CEO jobTitle founder nationality alumniOf owns mentions directed alumniOf alumniOf Math & Engineering provider Strategy & Innovation provider knowsAbout Rowing https://www.wikidata.o rg/wiki/Q159354 https://en.wikipedia. org/wiki/Rowing kg:/m/06f41 sameAs
3. • The Evolving Value of Schema Markup • How to Build a High Quality Content Knowledge Graph • What’s Required for Agent-Readiness? Agenda
4. Google Search Central Live Dubai Oct ‘25 Source: Search Engine Roundtable, Oct 23, 2025, Google: Links, Site Moves & Technical SEO Don't Fix Quality Issues “Structured Data is critical for modern search features” M
5. Schema Markup builds a Content Knowledge Graph from your content
6. What is a Content Knowledge Graph? A collection of relationships between the entities defined within the content on your website using a standardized vocabulary, from which new knowledge can be gained through inferencing.
7. Schema Markup Content Knowledge Graph Content Optimization Topic Authority, Entity Insights, Content Opportunities Content AI efficiencies through reusable data layer: MCP or NLWeb. AI / Innovation Entity Linking Taxonomy SEO Rich Results, Non Branded Queries, AI Overview Citations
8. Gartner names Knowledge Graphs as a “Critical Enabler” technology for Generative AI Gartner: 30 Emerging Technologies That Will Guide Your Business Decisions (Feb 12, 2024)
9. Source: Data.world (2023, November 13). A Benchmark to Understand the Role of Knowledge Graphs on Large Language Model Accuracy Data.world: Knowledge Graphs provide Higher Accuracy for LLM Responses in Enterprises by 300%
10. “Knowledge Graphs are no longer ‘nice to have’ — they are becoming a foundational layer for trustworthy AI.“ Mark van Berkel, Semantic Ontologist CTO @ Schema App
11. “The Semantic Web will enable machines to comprehend semantic documents and data, and enable software agents roaming from page to page to execute sophisticated tasks for users.” Tim Berners-Lee, Ora Lassila, and James Hendler The Semantic Web, Scientific America, 2001
12. What is the Agentic Web? It’s the Semantic Web in motion. It’s when AI-powered agents don’t just look up information, but actually do things for us — coordinating tasks, connecting services, and making recommendations on behalf of your customers.
13. Google’s Move Toward Agentic AI in Shopping • Google’s AI features rely on the Shopping Graph, built from product feeds, merchant data, and structured data. • High-quality structured data ensures your products are understood and accurately represented in Google’s agentic experiences. Source: Google, Nov 13, 2025, Let AI do the hard parts of your holiday shopping
14. Agentic Commerce Protocol Developed by OpenAI & Stripe An open standard model for enabling structured, secure commerce flows between buyers, AI agents, and businesses.
15. Microsoft’s CTO, Kevin Scott shares the future potential of the agentic web and the need for open protocols, ie. NLWeb. Source: Youtube, May 22, 2025, Microsoft Build 2025 Keynote | Kevin Scott Microsoft CTO Building the Agentic Web
16. NLWeb • Goal: Make it easy for any web publisher to create an intelligent, natural language experience for their site. • Leverages semi-structured formats like Schema.org, RSS and other data that websites already publish. • Created by RV Guha, who created Schema.org. Source: Microsoft, May 19, 2025 – Introducing NLWeb: Bringing conversational interfaces directly to the web Example of NLWeb on the Schema App Website
17. Guha created NLWeb to be the open standard for agents to discover endpoints on websites and interact with the data. • “No hallucinations because it uses Schema Markup” – RV Guha • Schema App is using NLWeb for our onsite search to learn more how it can be used to prepare for the agentic web. RV Guha Creator of NLWeb
18. Your Content Knowledge Graph Is the Bridge Between Your Brand and AI Understanding.
19. “Schema is a type of code that helps search engines and AI systems understand your content. Source: Microsoft, October 8 2025, Optimizing Your Content for Inclusion in AI Search Answers | Microsoft Advertising Microsoft shares how to prepare for AI “While there’s no secret strategy for being selected by AI systems, success starts with content that is fresh, authoritative, structured, and semantically clear.” Good News: Your content knowledge graph provides AI with a contextual understanding of your content.
20. How to Build a High Quality Content Knowledge Graph
21. High Quality Schema Markup Checklist Define: Use the right Schema.org Type to categorize your page Depth: Use relevant Schema.org properties that describe the content visible on your page Breadth: Implementing Schema Markup on all key pages on your website Connected: Use Entity Linking to describing the content and entity relationships on your key pages within your Schema Markup
22. Define: Use the right Schema.org Type to categorize your page SoftwareApplication
23. Leverage the Schema.org properties to describe your content in detail
24. creator > Organization name video review Depth: Describe content using Schema.org properties in Schema.org so it defined with greater detail description subjectOf > FAQPage
25. Breadth: Implement Schema Markup on all key pages on your website SoftwareApplication Organization ProfilePage
26. SoftwareApplication Connect: Use Schema Markup to connect entities across your key pages and the web creator Organization ProfilePage mainEntity > Person founder
27. Machines need to be able to infer things to answer questions How? By understanding the relationships between things (defined entities) owns mentions directed
28. What is an Entity? An entity is a thing with specific attributes and relationships to other entities
29. Entity Linking Entity Linking identifies entities in your content and links them to: • External authoritative sources (Wiki, Google) • Internal entities from your Content Knowledge Graph "mentions": { "@type": "Thing", "name": "knowledge graph", "sameAs": "http://g.co/kg/g/11jtypdlnf", "sameAs": "https://en.wikipedia.org/wiki/Knowledge_graph", "sameAs": "http://www.wikidata.org/entity/Q33002955", }
30. • Used Entity Linking to win local search & future-proof their SEO • 25%↑ in clicks • 30%↑ in impressions for non-branded queries within months of strengthening entity linking coverage Success Story
31. • Took control of their brand in search and boosted product visibility • 69%↑ in clicks • 116%↑ in impressions for non-branded product-related queries Success Story
32. Entity Optimization Audit Entity location, frequency, coverage Identify Identify entities in content Optimize Entities for accuracy Analyze Entity performance for actionable insights
33. Curating your entities controls the narrative Drives Results Mitigates Risk Aligns with Brand
34. What’s Required for Agent-Readiness?
35. • AI systems and browsers that take actions on your behalf – not just show results. • Websites become data sources and APIs for agents, not just destinations for people. • New standards (like MCP and NLWeb) are emerging to connect agents to your data safely and reliably. Enter the Agentic Web
36. Knowledge Graph • Accessible • Correct • Complete Agentic Entry Points Registry of actions into your business AI Governance • Trust • Accuracy • Compliance • Open Standards Agent Ready
37. Microsoft NLWeb • Turns a website into an AI-powered application with conversational interfaces. • Every NLWeb instance is also an MCP server, exposing site content to agents in a structured way. How Agents Can Talk to Your Data Model Context Protocol (MCP) • Open standard for connecting AI agents to external data sources. • Used by OpenAI, Anthropic, Microsoft, others – one integration, many agents.
38. Turning Your Graph into Agent Fuel Content Optimization Stack Published web content Add server-side schema markup Content Knowledge Graph Add high-quality graph features with curated external entities, topic modeling, and interlinking of entities Agentic Access Layer Securely exposes your graph to AI agents. • Access Controls, Governance, Monitoring • Model Context Protocol 1 2 3 4 Agentic Outcomes • Accurate, brand-aligned answers in LLMs • Controlled presence in agentic surfaces • Reduced hallucinations and mis-interpretations
39. Agentic Web END USERS Agentic Runtime Layer WEBSITE CONTENT CoPilot Agent Gemini Built-In Tools Search Indexes Browser • Agents will access your website directly. • The question is whether you control that experience. • MCP and NLWeb help you get ready.
40. Add Published Integrations + Tools END USERS Agentic Runtime Layer WEBSITE CONTENT CoPilot Agent Gemini Built-In Tools Search Indexes Browser • Define how agents should interact with your site. • Integrations route into your MCP server. • Content KG is the authoritative source behind the server. Published Integrations User Tools MCP Server Content KG
41. NLWeb END USERS Agentic Runtime Layer WEBSITE CONTENT CoPilot Agent Gemini Built-In Tools Search Indexes Browser • Build in conversational interface. • NLWeb is also an MCP server, gives agents a structured / standard way to consume your data. • Now the open web can talk back. Published Integrations User Tools MCP Server Content KG NLWeb UI NLWeb MCP
42. Controlled Experience END USERS Agentic Runtime Layer WEBSITE CONTENT CoPilot Agent Gemini Built-In Tools Search Indexes Browser • Published Integrations with NLWeb and MCP. • Content KG behind it all grounding with real data. • Agents don’t guess, they follow your instructions. Published Integrations User Tools Content KG NLWeb UI NLWeb MCP
43. AI Governance
44. Source: Gartner, February 17, 2025 – The Top CIO Challenges, According to 12k+ of Your CIO Peers “ Without AI-ready data foundations and practices, CIOs will be unable to deliver value from AI investments. In fact, most CIOs struggle to create trusted data foundations (i.e. data governance, data literacy, greater data collaboration) to enable AI-driven business outcomes.
45. Most AI vendors are still improvising. Schema App is already compliant with the standards that will define the next decade.
46. • Your website becomes the authoritative semantic source • Verified entities and definitions eliminate misinterpretation • Brand-aligned meaning stays consistent across all AI touchpoints • AI answers start from reality, not hallucinated context AI that represents your brand accurately because it’s built on governed, unambiguous truth. Trust AI grounded in your authoritative content
47. • Graph-grounded answers reduce hallucinations • Explicit triples enable explainability and traceability • Entity Management enforces semantic consistency across your entire footprint • MCP delivers structured, factual context at generation time • Proven accuracy gains (e.g., KG grounding: 91% vs. GPT-4 baseline 43%) Answers you can trust because they’re generated from real, governed facts. Accuracy Factual, consistent, explainable answers
48. Hallucinations and inconsistencies in LLMs are a feature Evidence shows that Knowledge Graphs (KGs) provide the foundation for more accurate, explainable, and trustworthy AI. What the research says: • Data.world: KGs provide 300% Increase in Accuracy for LLM Responses in Enterprises • LinkedIn (SIGIR ’24): KG-augmented customer service bots improved accuracy by 78% and cut resolution time by 29%. • Microsoft GraphRAG: Outperformed baseline RAG by providing better grounding and reasoning with graph structures. • Survey on Hallucination Mitigation: Academic reviews show structured data significantly reduces hallucinations in LLMs. • NUS + Cambridge (OKGQA 2025): Demonstrates KG grounding reduces hallucinations in open-ended QA benchmarks. • COLING 2025 (KG-FPQ): Benchmarking with 178k KG-derived false premise questions stress-tests factuality and trustworthiness.
49. Avoid proprietary traps and ensures long-term durability Your knowledge graph should be built on these web standards: • RDF + Schema.org semantics = universal machine readability • PROV-O for portable provenance • MCP + NLWeb provide open, inspectable access for agents Knowledge remains re-usable across search, AI, assistants, and internal tools Open Standards Interoperable, Portable, Future-proof
50. Agentic Entry Points Make your business callable by AI agents Organizations need a machine-readable registry of their business actions so agents know how to engage. 1 Identify Your Key Actions What customers come to you to do — Book, Buy, Apply, Request, Schedule. 2 Clarify the Customer Inputs What information an agent must collect to complete the action. 3 Partner With IT / Service Teams for the Technical Specs • Rules & conditions (eligibility, pricing, timing) • Authentication requirements • Success/failure definitions • EntryPoints: APIs, URLs, or workflows 4 Maintain a Central Registry A single, updated list of all actions and how they can be triggered.
51. Knowledge Graph • Accessible • Correct • Complete Agentic Entry Points Registry of actions into your business AI Governance • Trust • Accuracy • Compliance • Open Standards Agent Ready
52. What To Do Next
53. What Marketers Can Do Now Lock in the foundation ▢ Ensure consistent Schema Markup on key templates ▢ Deploy entity linking + topic taxonomy across priority content Expose your graph to agents safely ▢ Implement MCP Server as your AI-ready data source ▢ Pilot internal agent use-cases (e.g., support, sales enablement) before going public Build and tune your Content Knowledge Graph ▢ Consolidate duplicate entities ▢ Fill coverage gaps on strategic topics ▢ Align entities with business priorities
54. Schema Markup Content Knowledge Graph Content Optimization Topic Authority, Entity Insights, Content Opportunities Content AI efficiencies through reusable data layer: MCP or NLWeb. AI / Innovation Entity Linking Taxonomy SEO Rich Results, Non Branded Queries, AI Overview Citations
55. Questions? Download our FREE eBook ‘Mastering AI Search: Essential Strategies and Insights’ to find out how your brand and website can thrive in the era of AI search. Scan the QR code or visit https://bit.ly/tech- seo-connect-ai- ebook to download your free eBook!
56. Schema App is a scalable End-to-End Schema Markup and Content Knowledge Graph Solution for Enterprise Marketing teams. We help organizations use Schema Markup to build Content Knowledge Graphs to stand out in search, gain content insights and accelerate AI initiatives, through our industry-leading platform and Customer Success. Trusted by Enterprises Worldwide