What is structured data?
Structured data is machine-readable markup — almost always the schema.org vocabulary expressed as JSON-LD — that states what a page's content is: an Article, an Organization, a Person, a Product, a BreadcrumbList. Instead of leaving a machine to infer meaning from prose, you declare it.
It sits in the page's head or body as a small JSON block. It changes nothing a human sees — its entire audience is software.
What it actually does
Two proven jobs. First, eligibility for rich results: breadcrumbs, article treatments, and other SERP features only appear when the matching, accurate schema is present (presence makes you eligible — it doesn't guarantee display). Second, entity understanding: it disambiguates your organization and author and helps connect them in the knowledge graph.
This estate ships Organization, Person, BlogPosting, and BreadcrumbList JSON-LD — generated from canonical data so the markup can't drift from what's on the page. Those are the real, defensible uses.
What it does NOT do
It is not a proven AI-citation lever. There's no controlled evidence that adding schema lifts your odds of being cited in AI Overviews, ChatGPT, or Perplexity — citation tracks answer quality, not markup (see the cornerstone). And it actively backfires when misused: AggregateRating or Review schema on pages that aren't eligible for it is a documented demotion flag.
So the failure mode isn't too little schema — it's inaccurate schema bolted on to fake signals.
Which types are worth adding — and how do you test them?
Lead with the types that earn rich results or anchor your entities: Organization and Person (who you are, for the knowledge graph), BreadcrumbList (navigation in the result), Article/BlogPosting (article treatments), and Product, Recipe, or FAQPage only where the page genuinely qualifies. Skip speculative markup with no eligible display.
Then test it: Google's Rich Results Test shows which rich results a URL is eligible for, and the Schema Markup Validator catches syntax errors. Validating before you ship is how you stay clear of the inaccurate-schema failure mode — the markup either earns a real feature or it shouldn't be on the page.
How to use it well
Mark up what is genuinely on the page, accurately, using only eligible types; keep the markup in sync with the visible content; and never assert ratings, reviews, or facts the page doesn't actually carry. Generating it from the same source as the page — as we do — is the cleanest way to guarantee it never drifts into a false claim.
Accurate schema for its real jobs, paired with genuinely good content, is part of our Search & AI visibility practice — and it pairs with shipping llms.txt as hygiene rather than as a lever.