JSON-LD Schema Validator

Structured data fails quietly. A trailing comma, a missing @type, or a required property left out does not break the page - it just means the rich result never appears and nobody tells you why. This validator makes those failures visible before deployment.

It works two ways. Paste a JSON-LD block or a full script tag and it parses and checks it instantly in the browser. Or give it a URL, and the server fetches the page HTML, extracts every application/ld+json block on it, and validates each one separately with its own error and warning counts.

The URL mode is the one that catches real-world problems, because it validates what is actually being served rather than what you think you wrote. Themes, plugins, and consent scripts all inject their own JSON-LD, and a page frequently carries three or four blocks that were never reviewed together.

Errors are structural: the JSON does not parse, @context is absent or wrong, @type is missing, or a property that the type genuinely requires has no value. A block with an error will not produce a rich result, and often will not be understood as structured data at all.

Warnings flag recommended properties whose absence narrows what a search engine can display - an Article without an author or datePublished, a Product with no offers block, a Recipe missing cook time. The markup is valid, but you are leaving display features on the table. Fix errors before deploying; treat warnings as a backlog.

A well-marked-up article page might carry Organization and WebSite markup from a site-wide template, BreadcrumbList from the navigation component, Article from the post template, and FAQPage from a plugin rendering the questions at the bottom. All four coexisting is correct and expected.

The failure mode is contradiction rather than quantity. Two Organization blocks with different names, a BreadcrumbList whose final item does not match the page URL, or an FAQPage describing questions that were removed from the visible copy last quarter. The accordion view groups results by block precisely so you can read them side by side and notice when two of them disagree.

Google requires that structured data describe content actually visible on the page. FAQ markup for questions a visitor cannot see, or Review markup for ratings that appear nowhere, is a guidelines violation and a common cause of manual actions against rich results.

In paste mode, the input is scanned for script tags of type application/ld+json and their contents extracted; raw JSON with no wrapper is accepted directly. Each block is JSON-parsed, and a parse failure surfaces immediately with the position reported by the parser. Arrays and @graph containers are unpacked so each entity inside is analysed as its own block.

For every entity the validator confirms that @context resolves to schema.org, that @type is present and is a recognised type string, and that the required properties for that type are present and non-empty. Recommended-but-optional properties generate warnings. Each block is then re-serialised with consistent indentation so you can copy a corrected version straight out of the result panel.

In URL mode the request goes to a server endpoint that fetches the page HTML and extracts every application/ld+json script it finds, reporting both the number of script tags and the number of entities inside them - those numbers differ whenever a single script carries an array. Only publicly reachable pages work; anything behind authentication, a login wall, or aggressive bot protection will return a fetch error rather than results.