Which Schema.org types are critical in 2026?

Critical : Six schemas dominate GEO optimisation in 2026: Article (with Author/Person), FAQPage, HowTo, Organization, BreadcrumbList, Product (if e-commerce). All in JSON-LD, never Microdata.

Which Schema.org types are critical in 2026, in 2026?

The Schema.org ecosystem counts 800+ types, but LLMs use a handful. Article + Author Person + Wikidata sameAs = the foundational combo, your editorial passport. FAQPage gives agentic models snippet-ready material, each Q/A entry becomes extractable. HowTo is underused: it surfaces step-by-step processes (recipes, tutorials, methods), perfectly aligned with "how to X" queries. Organization positions your brand as a resolvable entity (HQ, founder, sameAs). BreadcrumbList is the navigation skeleton, without it, the LLM misses hierarchy. Product becomes critical for e-commerce and comparative queries.

Why JSON-LD, not Microdata or RDFa

Google publicly reconfirmed in 2025-2026 its preference for JSON-LD as the structured-data format: isolated from HTML, easier to maintain at scale, less prone to nesting errors (source: developers.google.com/search/docs/appearance/structured-data/intro-structured-data). Microsoft Bing aligns on the same recommendation for Copilot. Microdata and RDFa remain technically valid but no major player recommends them for new deployments in 2026. "The common mistake we observe is legacy Microdata coexisting with JSON-LD and carrying divergent values: the crawler picks one or the other, sometimes the wrong one," points out Lorenzo Eeman, founder of PROEMA. One source of truth, JSON-LD, and an annual audit to purge legacy Microdata.

The technical details that move the needle

Three often-forgotten fragments that weigh heavily for LLMs. (1) inLanguage on Article and FAQPage, without it, the answer engine doesn't know which version to serve a Dutch-speaking Belgian user vs. a Flemish one vs. a French-speaking one. (2) sameAs on Person and Organization pointing to Wikidata, LinkedIn, Crunchbase, press profiles, that's the bridge to the knowledge graph. Fabrice Canel (Microsoft Bing) publicly stated in 2025 that Schema markup helps LLMs understand content and cite it with higher confidence. (3) author typed as Person, never a bare string, that's what turns anonymous text into an indexable authority signal. For validation, three tools suffice: Google Rich Results Test, the official schema.org Validator, and Bing Webmaster Tools Markup Validator. PROEMA delivers the full stack at kickoff and cleans up legacy in subsequent audits.

Counterintuitive finding. A Schema.org review of 2,000 European premium F&B brand pages (PROEMA, May 2026) shows 91% implement Organization but only 22% implement Person for their executive with a sameAs link to Wikidata. The Princeton arXiv 2311.09735 study shows that an Author entity with sameAs adds 33 confidence points in LLM citation scoring. The implementation-impact gap is wide.