SeenandCited

Schema.org for AI Visibility

Master the specific Schema.org types and properties that enhance Large Language Model (LLM) comprehension and entity linking for AI-driven search results.

12 min read
Foundations

Introduction to Schema for AI Visibility

In the era of Generative Search Experiences (SGE) and AI Overviews, Schema.org structured data has evolved from a rich-snippet tool into a primary source of ground truth for Large Language Models (LLMs). While traditional SEO often focuses on generating star ratings in SERPs, the AI Visibility Practitioner focuses on Entity Disambiguation. For an AI to recommend your client as a solution, it must first understand exactly what that client is, who they serve, and how they relate to other verified entities on the web. This lesson provides a tactical guide to the schema types that move the needle in AI visibility.

The Shift: From Search Engines to Answer Engines

Traditional search engines used structured data to display visual enhancements. AI answer engines use it to build a Knowledge Graph. When an LLM crawls a page, it looks for explicit markers to confirm its internal probability models. For instance, if an AI is 70% sure a page refers to 'Apple' the tech company rather than 'Apple' the fruit, the presence of Organization schema with a sameAs link to a Wikidata entry increases that confidence to 99%. High confidence leads to inclusion in AI citations.

The Hierarchy of Importance

Not all schema types are created equal for AI visibility. We categorise them into three tiers:

  1. Core Identity: Who you are (Organization, Person).
  2. Contextual Authority: What you know (Article, FAQPage, HowTo).
  3. Transactional Proof: What you offer (Product, Service, Offer).

Core Identity: Building the Knowledge Graph

The foundation of AI visibility is the Organization or Brand schema. For an AI to cite you as an authority, it must reconcile your website with real-world entities.

Key Properties for Identity

  • @id: Use a unique URI (usually the homepage URL) to serve as the global identifier for the entity across the site.
  • sameAs: This is the most critical property for AI. Link to official social media profiles, but more importantly, to authoritative database entries like Wikidata, Wikipedia, or Crunchbase.
  • knowsAbout: List specific topics or keywords that the entity is an expert in. This helps the AI map the organization to specific niche clusters.
  • memberOf: Links the entity to industry bodies or trade associations, providing third-party validation of status.

Contextual Authority: Supporting AI Summarisation

AI Overviews frequently synthesise information from multiple articles. Structured data helps the AI parse the most important parts of your content quickly.

Article and NewsArticle

  • author: Use a Person object rather than a string. Include the author's jobTitle and sameAs links to their professional profiles to satisfy E-E-A-T requirements.
  • keywords: Explicitly state the primary entities discussed in the article.
  • significantLink: Point to the primary external resources that support the article's claims.

FAQPage

While Google has reduced the visibility of FAQ snippets in standard search, LLMs still use FAQPage schema to extract direct answers for conversational queries. Ensure each Question and AcceptedAnswer is concise and factual.

Transactional Proof: Enhancing Product Discovery

For e-commerce and B2B services, the Product and Offer types are vital. AI models like ChatGPT with Search or Google's shopping graph rely on structured data to compare prices, availability, and features.

Beyond the Basics

  • audience: Define who the product is for (e.g., 'Target Audience: SEO Professionals').
  • isRelatedTo: Link to complementary products to help the AI understand where you sit in the market ecosystem.
  • review: Aggregate reviews provide the 'social proof' indicators that AIs often cite when explaining why they recommended a specific result.

Worked Example: Optimising a SaaS Company

Let’s look at a hypothetical SaaS company, 'DataViz Pro'.

Standard Schema: Only uses Organization with a name and logo.

AI-Optimised Schema:

  1. Organization: Includes sameAs links to their LinkedIn and a Wikidata entry for 'Data Visualization Software'.
  2. Founder: A Person schema for the CEO, linked to their published research on IEEE using knowsAbout.
  3. SoftwareApplication: Includes featureList detailing specific AI integrations and operatingSystem requirements.
  4. HowTo: A guide on 'How to visualize Big Query data' with step-by-step schema, making it a prime candidate for an AI-generated 'how-to' summary.

Common Implementation Pitfalls

Avoid these mistakes that confuse LLMs:

  • Schema-Content Mismatch: Providing structured data for information not visible on the page. This is viewed as a manipulation attempt by both traditional algorithms and AI filters.
  • Broken Nesting: Failing to nest entities (e.g., placing the Author inside the Article rather than as a separate, unlinked entity). Proper nesting shows the relationship clearly.
  • Generic @types: Using Thing when you could use a specific subtype like FinancialService or MedicalBusiness.

Putting it into Practice

To implement an AI-first schema strategy, follow these steps:

  1. Entity Audit: Identify the 5 most important entities associated with your client (Brand, CEO, Lead Product, Primary Location, Core Topic).
  2. Wikidata Mapping: Find the corresponding Wikidata IDs for these entities to use in sameAs properties.
  3. Author Verification: Update all blog posts to use Person schema for authors, including alumniOf and knowsAbout properties.
  4. Nesting Check: Ensure your WebPage schema contains the About (main topic) and Mentions (secondary topics) properties to define the page's focus for the AI.
  5. Validation: Use the Schema Markup Validator (schema.org) rather than just the Rich Results Test. The Rich Results Test only shows what Google displays, while the Validator shows what the AI sees.

Visual diagram

[ diagram placeholder ]

A workflow diagram showing an AI crawler processing a webpage: one path shows unstructured text being processed with 60% confidence, while the other shows Schema.org data providing 99% confidence for entity linking.

Exercise

Take a client's 'About Us' page and identify three potential 'sameAs' URLs from authoritative sources (Wikidata, LinkedIn, or an industry directory). Then, write a JSON-LD 'Organization' snippet that nests these sources and defines the 'knowsAbout' property with three core topics.

Key takeaways

  • Schema.org is a primary source of ground truth for LLMs and AI answer engines.
  • Entity disambiguation via 'sameAs' links to Wikidata is crucial for AI confidence.
  • The '@id' property creates a unique global identifier for your brand in the Knowledge Graph.
  • Use 'Person' schema for authors to satisfy AI-driven E-E-A-T evaluations.
  • The 'knowsAbout' property explicitly signals topical expertise to AI crawlers.
  • FAQPage schema remains a high-value source for conversational AI responses.
  • Specific subtypes (e.g., 'SaaS' instead of 'Thing') provide clearer context for AI models.
  • Nesting items (e.g., Author within Article) demonstrates the relationship between entities.
  • AI models use 'Product' and 'Offer' data for comparative analysis in shopping queries.
  • Validation should focus on semantic accuracy, not just visual rich snippet eligibility.

Lesson Quiz

Pass at 70%.

1. Which Schema.org property is most effective for linking a website entity to its established entry in a database like Wikidata?
2. What is the primary benefit of using the '@id' property in your JSON-LD?
3. Why should you use 'Person' schema for authors rather than a simple text string?
4. Which property would you use to tell an AI which specific topics an Organization is an expert in?
5. True or False: The Schema Markup Validator is preferred over the Rich Results Test for AI Visibility auditing.
6. In the context of AI, what does 'Entity Disambiguation' mean?
7. Which property is most useful within a 'WebPage' type to signal the primary subject matter to an AI?
8. What happens if your schema says one thing but the visible page content says another?
9. Which schema type should be used for a B2B software offering to help AI comparison tools?
10. How do AI answer engines typically use FAQPage schema?
Create a free account to save progress and earn a certificate.
Auditing Your Brand EntityBack to moduleEntity Linking Across the Web