How Structured Data (JSON-LD) Affects LLM Citations: Not Just Google, OpenAI Reads It Too

What Most People Get Wrong About JSON-LD

For the past decade, SEO consultants have taught this:

“Add JSON-LD schema so Google’s SERP can show star ratings, FAQ accordions, product prices, and other rich results.”

This claim is not wrong in itself, but it only tells half the story. Google rich results are an “added bonus,” but the real strategic value of JSON-LD only surfaced in the AI era: LLMs parse this markup heavily during training.

Why Do LLMs Care About schema.org?

Before training, LLMs must first “structure” web page text—distinguishing the title, body, author, date, and other elements. Traditional HTML parsing relies on semantic elements like H1 / H2, <article>, and <time>, but these have limited credibility (anyone can misuse H1).

What JSON-LD provides is facts at the level of an “author’s own declaration”:

• "@type": "Article" — I declare this page is an article
• "author": {"@type": "Person", "name": "Dr. Lee"} — the author is this person
• "datePublished": "2024-03-15" — the publication date
• "about": {...} — the topic entity

During training, LLMs treat these declarations as structured facts written into the vector space. The same content with vs. without JSON-LD produces very different “understanding” by the AI.

The 4 Schemas With the Biggest Impact on GEO

1. Organization (Company Entity)

Every brand’s official site should have this on its homepage and “About Us” pages. It includes:

• name — the full company name
• url — the official website
• logo — the logo image
• sameAs — Wikipedia, LinkedIn, and official social media links
• contactPoint — customer service information

Why it matters: It establishes the authoritative foundation of a “brand entity.” Without Organization schema, an LLM sees only a string of text, with no structured concept of a “company.”

2. Article (Article Content)

Every blog / news article should have this. It includes:

• headline — the title
• author — the author (with Person schema)
• datePublished / dateModified — the dates
• publisher — the publishing entity
• image — the main image

Why it matters: It directly affects the E-E-A-T dimension score. To an LLM, an article without Article schema has heavily discounted authority.

3. FAQPage (Q&A Structure)

If your content is in FAQ format, you must add FAQPage schema. It includes:

• mainEntity — the array of Q&As
• each question with a Question and an Answer

Why it matters: Perplexity and Google AI Overviews cite FAQ structures at extremely high rates. An FAQ that uses H2s for questions with plain-text answers vs. an FAQ wrapped in FAQPage schema differ by more than 5x in citation probability.

4. Product / Service

E-commerce and SaaS should add this. It includes:

• name, description, brand
• offers — price, purchasable regions
• aggregateRating — ratings
• review — reviews

Why it matters: When a user asks an AI for “recommendations for XYZ,” the LLM internally compares the structured attributes of candidate products. Product information that isn’t structured gives the AI nothing to compare against.

Why JSON-LD > Other Formats?

In the past there were also two structured-data formats: Microdata and RDFa. Why does GEO prefer JSON-LD?

In its official 2024 documentation, Google formally recommended JSON-LD as the primary format. OpenAI and Anthropic also prioritize parsing JSON-LD during training (other formats are sometimes ignored or misinterpreted).

Common Mistakes

Things we’ve seen in client audits:

• Mistake 1: Going live without ever testing it through a schema validator. When the schema has errors, Google / LLMs ignore the entire block outright.
• Mistake 2: FAQPage schema content that doesn’t match the page’s actual Q&A—LLMs detect this inconsistency and penalize the entire website.
• Mistake 3: Placing Organization only on the homepage but not on subpages—when LLMs parse a subpage, they still can’t see the company entity.

What an Audit Can Tell You

GeoWeb’s “Structured Data” dimension (10% weight) lists, item by item:

• which schemas your site has and which it’s missing
• whether the schema has syntax errors
• whether the schema content matches the page’s visible content
• whether there is any “fake schema” (content that doesn’t exist but is declared as existing in the schema)

👉 Free GEO Audit

If, after reading the report, you need technical help adding and maintaining schema, we offer a GEO consulting service: contact@geoweb.tw

GEO Advanced Series #9. Previous article: “Why E-E-A-T Matters More in the AI Era Than Ever Before”