Pick a schema type, fill in the fields, and get valid JSON-LD ready to paste into your page. Nine of the types that actually earn rich results are covered — Article, FAQ, Product, Organization, Local Business, Breadcrumbs, Website, Event and Recipe — each with its required fields marked and checked before you copy. Nested entities are built properly rather than flattened: an author becomes a Person, a price becomes an Offer, an address becomes a PostalAddress. That distinction is what separates markup that validates from markup that actually produces a rich result, and it is where hand-written JSON-LD most often goes wrong. Empty fields are dropped entirely, because a present-but-blank property is treated as malformed rather than simply absent. One caveat worth stating plainly: only mark up content that is genuinely visible on the page — marking up things a visitor cannot see is a policy violation, not a shortcut.
| Type | Use it on | Key required fields |
|---|---|---|
| Article | Blog posts, news | Headline, author, date published |
| FAQPage | Pages with visible Q&A | Question and answer pairs |
| Product | Product pages | Name, price, currency |
| Organization | About or home page | Name, URL |
| LocalBusiness | A business with an address | Name, street, city |
| BreadcrumbList | Any page in a hierarchy | Ordered name and URL steps |
| WebSite | Home page | Name, URL |
| Event | Event listings | Name, start date, venue |
| Recipe | Recipe pages | Name, ingredients, steps |
JSON-LD is structured data expressed as a JSON block in a script tag, separate from your visible HTML. Google explicitly recommends it over microdata and RDFa because it does not entangle your markup with your content, which makes it far easier to maintain and much harder to break when the page layout changes.
Anywhere in the page — head or body both work. Head is conventional. You can have several JSON-LD blocks on one page, which is the normal way to combine, say, Organization and BreadcrumbList markup.
No, and nothing can. Valid markup makes a page eligible; whether Google shows a rich result depends on page quality, the query, and its own judgement. Validating correctly is necessary but not sufficient — and FAQ rich results in particular have been heavily restricted in recent years.
No. Structured data must describe content a visitor can actually see. Marking up FAQs that are not rendered, or reviews that do not exist, is a structured data policy violation and can result in a manual action against the site. Use it to describe what is there, not to claim more.
Run the output through Google's Rich Results Test and the Schema Markup Validator. This tool builds structurally correct JSON-LD, but only Google can tell you whether a specific page qualifies for a specific rich result.
No. The JSON-LD is assembled in your browser. Business details, prices and unpublished content you enter are never sent to a server, logged, or stored.
Each schema type carries a bundled field definition describing its properties, which are required, and how they nest. The JSON-LD object is assembled from those fields and then compacted: any property whose value is empty is removed entirely rather than emitted as an empty string or null, because Google's structured data parser treats a present-but-empty property as malformed rather than absent. Nested types are constructed as real nested objects with their own @type — an author becomes a Person, a product's price becomes an Offer, an address becomes a PostalAddress — since a flat string where the vocabulary expects an entity is the most common reason markup validates as a warning rather than an error but still fails to produce a rich result. Availability values are prefixed with the schema.org URL, breadcrumb positions are numbered from one in the order given, recipe steps become HowToStep objects, and multi-line inputs are split and trimmed into arrays. Required fields are checked and reported before you copy, and the output is serialised with JSON.stringify so quotes and newlines inside text are escaped correctly. Everything runs locally.