What structured data does for AI answers
Engines can extract facts from prose, but extraction is lossy. JSON-LD tells them directly: this is our official name, this is the founding date, this page answers these questions, this content was updated on this date. Structured data converts your most important facts from probabilistically extracted to declared — and declared facts win ties.
The types that earn their bytes
| Type | Job | Where |
|---|---|---|
| Organization / LocalBusiness | Entity card: name, logo, sameAs, contact, founding facts | Homepage + about |
| Article / TechArticle | Authorship, datePublished, dateModified — freshness and E-E-A-T signals | Every content page |
| FAQPage | Question-answer pairs in liftable form | Pages with visible FAQs — markup must match visible content |
| Product / Offer | Names, specs, price ranges, availability | Product pages |
| BreadcrumbList | Site structure and topical grouping | Every page |
| DefinedTermSet | Glossary entries as citable definitions | Glossary/wiki hubs |
Rules that keep markup honest
- Markup mirrors the visible page. FAQPage schema for questions not shown on the page is a spam signal, not a hack.
- Dates are real. dateModified that bumps without content change teaches engines to ignore your dates.
- Validate. Run the Rich Results test and a JSON-LD parser; broken markup is worse than none.
- sameAs discipline. Point only to profiles you control or that are authoritative — a wrong sameAs actively merges you with the wrong entity.
Frequently asked questions
Does schema markup directly improve AI citations?
Indirectly but materially. Markup clarifies entity facts, dates and structure that grounding systems otherwise extract imperfectly from prose. Think of it as removing ambiguity rather than adding ranking fuel — the citations follow from being correctly understood.
Which pages need JSON-LD most?
Homepage and About (Organization), every article (Article with dates and author), every page with visible FAQs (FAQPage), product pages (Product/Offer). BreadcrumbList everywhere is cheap structure. Glossaries get DefinedTermSet.