Structured Data

Prose forces machines to infer; structured data lets you declare. A LocalBusiness object with name, address, phone, and hours; an Article with author and dates; a Dataset describing published research — each is a claim stated in the format machines parse first.

For a local service business the useful set is short. LocalBusiness or a more specific subtype (Dentist, Plumber, LegalService) carrying name, address, phone, hours, and service area. Service objects for what you actually sell. Article with author and dates on anything editorial. BreadcrumbList so the site's structure is explicit. Organization with sameAs links to the profiles that corroborate your identity. That list covers nearly everything worth marking up, and adding types beyond it rarely changes anything.

Two honest caveats. Markup describes content, it does not improve it — schema on a thin page is a label on an empty box. And markup types are not ranking bonuses; Google retired FAQ rich results in May 2026, so FAQ markup specifically no longer earns a special display.

The FAQ deprecation is a useful lesson about where the value actually sits. Sites that added FAQ schema to chase a rich result lost the benefit overnight. Sites that added it because their pages genuinely answered questions kept everything that mattered, because the answers were the asset and the markup was a description of them. Any markup strategy that survives a display change is one built on real content.

For the AI layer the value is different from rich results and more durable: unambiguous facts. An assistant reading a page with declared hours, address, and services does not have to infer them from prose that may be ambiguous or out of date. Accurate, minimal markup that matches visible content is the standard worth holding — see JSON-LD for the implementation discipline that keeps the two in sync.