Structured data is a translation layer, not a ranking hack
Every developer has sat through a marketing meeting where schema markup was described as the thing that will "make Google understand the site". That framing is wrong, and it leads to bloated JSON-LD pasted everywhere with no plan. Schema is simpler than the hype: it is a machine-readable summary of facts the page already states, written in a vocabulary — schema.org — that search engines and AI systems can parse without guessing.
Done properly, it buys eligibility: rich results, cleaner knowledge-panel data, and content that answer engines can lift accurately. Done badly, it buys warnings in Search Console and, in the worst case, a manual action for deceptive markup. The difference is discipline, not volume.
The types that earn their keep
Organization and WebSite
Every site should state who publishes it: name, logo, sameAs links to real profiles, and contact details that match the footer. This is the entity anchor everything else hangs from. AI assistants resolving "which company is this" lean on exactly these signals.
Article and BlogPosting
For editorial content, mark up the headline, author, dates and image honestly. The author should be a real Person entity with a name that matches the byline, not "Admin". Publication and modification dates must match what the page shows — a mismatch between the visible date and the structured date is the kind of inconsistency that erodes trust in the whole block.
FAQPage — with restraint
FAQ markup is legitimate when the questions and answers are visible on the page, word for word. It is not a place to hide extra content for crawlers. Google has steadily reduced how often FAQ rich results appear, but the markup still helps other systems parse question-answer content. One FAQ block per page, generated from the same source as the rendered HTML, never hand-maintained in two places.
Product, Service and LocalBusiness
Where they apply, these carry the facts users compare: price ranges, areas served, opening hours, ratings that genuinely exist on the page. Never mark up reviews you do not have. Ever. It is the fastest route to losing rich results entirely.
Implementation rules that prevent pain
Generate JSON-LD from the same data layer that renders the page. The moment a human edits markup separately from content, the two drift apart, and drift is what validators — and eventually penalties — catch. In a component-based stack, that means a single Schema component fed by the page's own props, serialised once in the document head or at the end of the body.
Validate in three places: the schema.org validator for syntax, Google's Rich Results Test for eligibility, and Search Console's enhancement reports a week after deployment for reality. Each catches a different class of error.
Keep the graph connected. Use @id references so the Article's author points at the Person, the publisher points at the Organization, and the WebPage ties them together. A connected graph is what lets a machine build one entity out of your pages instead of a pile of unrelated claims. A step-by-step walkthrough of this wiring, with examples for the common page types, is in this schema markup guide — worth an hour of any developer's time before a site-wide rollout.
What schema will not do
It will not fix thin content, slow pages or a site nobody links to. It will not guarantee rich results; it makes you eligible and lets Google decide. And it is not a substitute for saying the facts clearly in the visible content — increasingly, the systems consuming your pages compare the two, and the visible text wins ties.
The bottom line
Treat schema as documentation for machines: accurate, generated, connected and boring. Twenty correct properties maintained by your data layer beat two hundred pasted ones, every time.
A final operational point: assign ownership. Schema breaks silently — a template refactor drops the author property, a CMS update starts emitting a second FAQ block, a migration changes every @id. Add the enhancement reports to a monthly check, the same way you would review error budgets, and treat a sudden drop in valid items as a bug with a cause, not a mystery. The teams that keep their rich results for years are not the ones who wrote the cleverest markup on day one. They are the ones who noticed within a month when it broke.
Top comments (0)