Schema markup validator

The validator fetches the page you give it, pulls out every <script type="application/ld+json"> block, and checks each one against the property table below. It reports three kinds of finding and nothing else:

• Error — the block is not valid JSON, has no @context or @type, or is missing a property that is required for that type. A required property missing means the markup is not eligible for the rich result it is aiming at.
• Warning — a recommended property is missing. The markup still works; it says less.
• Pass — every property in the table is present.

There is no score. A score would need a weighting, and any weighting we put on "missing dateModified" against "missing headline" would be a number we made up. Each finding names the exact type and property instead, so you can fix it directly.

Limits worth knowing. The tool reads the HTML the server returns. Structured data injected by client-side JavaScript after page load will not be seen — which is also true of some crawlers, so a block that only exists after hydration is a real finding, not a false negative. Microdata and RDFa attributes are counted but not validated. Types outside the table below are checked for syntax only, and the report says so rather than guessing at their requirements.

Required and recommended properties follow Google's published structured data documentation for each rich result type. This is the whole rule set — nothing is applied that is not listed here.

A language model reading your page has to work out what is a fact, what is navigation and what is marketing. JSON-LD removes that guesswork: it states, in a machine-readable form, that this string is the author, this one is the publication date, this one is the price. When an assistant assembles an answer and needs to attribute a claim, unambiguous markup is the difference between being quoted with your name on it and being paraphrased without attribution.

The three that do the most work, in order:

• Organization — establishes who publishes the site. Without it, a model has no stable entity to attach your content to, and your brand name is just a string that happens to recur.
• Article or BlogPosting with author and dateModified — freshness and authorship are two of the signals that decide whether a page is used as a source or skipped for a newer one.
• FAQPage — question and answer pairs are the shape an assistant can lift directly. A well-marked FAQ answer is close to ready-to-quote text.

Markup does not make a weak page strong. It makes a strong page legible.

• Block is not valid JSON. Usually a trailing comma, a smart quote pasted from a document, or an unescaped quotation mark inside a string. The whole block is discarded when this happens — one stray character costs you all the markup on the page.
• No @context. Every top-level node needs "@context": "https://schema.org". Without it the block is arbitrary JSON, not JSON-LD, and is ignored.
• Missing author on an Article. Use a nested Person or Organization object with a name and, ideally, a url pointing at a real bio page. A bare string works but carries less.
• Missing dateModified. Emit it from your CMS, in ISO 8601. Do not hardcode it, and do not bump it on every deploy — a date that always says today tells a crawler nothing.
• Missing offers on a Product. Nest an Offer with price, priceCurrency and availability. Price without currency is not usable.
• Markup that contradicts the page. Not something this tool can detect, but worth stating: the values in your JSON-LD must match what a reader sees. Marking up a rating you do not display is a policy violation, not a shortcut.