JSON-LD

JSON-LD puts structured data in a self-contained script block rather than weaving attributes through the markup, which is what the older microdata and RDFa formats required. Google recommends it, and the separation is the reason: the data can be generated, validated and changed without touching the presentation.

For a build pipeline this is decisive. A block that is emitted from the same source as the page content cannot drift out of sync with it, whereas attributes sprinkled across a template survive exactly until someone restyles the component.

Multiple blocks on one page are fine, and are usually cleaner than one deeply nested object — a page can carry its main entity in one script and its breadcrumb trail in another without either becoming hard to read.