How JSON-LD works
Search engines read your page as text and infer what it is about. Structured data removes the inference by stating it outright: this is a local business, this is its address, this is an article, this is its author. JSON-LD is the way that statement is written — a self-contained block of script placed in the page’s code, marked as linked data, and completely separate from the words a visitor sees.
Inside the block, each thing you describe is an object with a type and a set of properties, and objects can refer to each other. A vocabulary, almost always schema.org, supplies the type names so that everyone means the same thing by “opening hours” or “author”. Because the block is independent of the layout, redesigning a page does not disturb it — which is the practical reason Google states a preference for JSON-LD over the older approaches that wrap attributes around visible HTML.
Why JSON-LD matters
It is how you become eligible for the richer results that appear beyond a plain blue link: review stars where genuine reviews exist, event dates, product availability, frequently asked questions, breadcrumbs, business details in a knowledge panel. Eligibility is not a guarantee — search engines decide when to show these — but without the markup the question never arises.
It has become more useful, not less, as answer engines and assistants have grown. A clearly described entity is easier for any machine to read confidently, whether it is building a search result or summarising a page, which is why structured data keeps appearing in conversations about AI-generated answers.
Common mistakes with JSON-LD
The serious one is describing something the page does not show. Marking up review ratings you did not collect, prices that are not listed, or an FAQ that no visitor can read is against Google’s guidelines and risks a manual penalty against the site. The markup is meant to describe the page, not to improve on it.
Then there are the quiet technical failures. A block copied from a generator and left with its example values still in place. Two competing descriptions of the same entity, one from a plugin and one from the theme, each contradicting the other. Markup injected by JavaScript that the crawler never executes, so a validation tool run in your browser passes while the search engine sees nothing. And a single missing bracket or a straight quote replaced by a curly one, which invalidates the entire block rather than one line of it.
How to act on it
Start with the types that genuinely describe your pages — the organisation, the local business, articles, services, products, breadcrumbs — and resist adding more for its own sake. Generate the values from the same source as the visible content wherever you can, so the two cannot drift apart when a price or an address changes.
Validate every template once with Google’s Rich Results Test and the Schema Markup Validator, then check the rendered page source rather than your draft, because plugins and tag managers both add markup of their own. Watch the enhancement reports in Search Console after launch, since that is where errors surface at scale. If several plugins are competing, turn all but one off; a single accurate description beats three arguing ones. Where the markup itself is the project, treat it as ongoing schema markup work that is reviewed whenever templates change.
