Structured data that actually gets picked up: what Google does with it, and what it ignores

Structured data gets treated as an SEO checkbox more often than it gets understood - add some JSON-LD, tick the box, move on. That approach usually produces markup that's technically present and functionally useless, because it either doesn't match what's on the page, uses the wrong schema type for the content, or misses the specific fields Google actually needs to build a rich result.

It's worth being precise about what schema markup does and doesn't do, because the gap between the two is where most implementations go wrong.

1. Schema doesn't improve rankings directly

This is the first misconception worth clearing up: adding JSON-LD is not a ranking factor in the way page speed or content quality are. What it does is make a page eligible for rich results - star ratings, FAQ accordions, breadcrumb trails, article cards - in search listings. Those rich results tend to improve click-through rate, which is a real, measurable traffic gain, but it's a different mechanism than ranking higher for the same query.

2. It must match the visible page exactly

Google's structured data guidelines are explicit about this and increasingly strict in enforcement: the content in your JSON-LD has to reflect what's actually visible on the page. FAQPage markup listing questions and answers that don't appear as visible text is a policy violation, not a grey area, and can get the markup ignored or the page penalized. The safest pattern is generating the schema from the same data source that renders the visible content, so the two can never drift apart.

3. Use the specific type, not the closest generic one

A common failure mode is reaching for a generic type - `Thing`, or a loosely-matched `CreativeWork` - when a specific one exists and unlocks a real rich result. A blog post should usually be `Article` or `BlogPosting`, not just `WebPage`. A services business should use `Service` for individual offerings, not bury everything inside one `Organization` block. The more specific and correct the type, the more of Google's supported rich-result features it becomes eligible for.

4. One entity, one @id, reused everywhere

Sites that repeat a full `Organization` object on every page - slightly differently worded each time - make it harder for Google to confidently connect the dots into a single entity. Assigning a stable `@id` to your core entities (the organization, the website) and referencing that `@id` from every other schema block - article authorship, service listings, breadcrumbs - signals a single, consistent entity far more clearly than repeating a similar-but-not-identical block on every template.

{
  "@type": "Article",
  "headline": "...",
  "author": { "@id": "https://example.com/#org" },
  "publisher": { "@id": "https://example.com/#org" }
}

5. Validate against the actual rich-result test, not just a JSON linter

• Valid JSON syntax is necessary but not sufficient - a schema.org type used with the wrong required fields still parses fine and still gets ignored.
• Run every template through Google's Rich Results Test, not just a generic JSON-LD validator, since it checks against the specific fields each rich result requires.
• Re-check after any template refactor - a field silently dropped during a redesign invalidates markup that used to work.
• Monitor the Enhancements reports in Search Console periodically; Google will flag markup that stops validating in production, often well after the change that broke it.

Structured data is a contract: you're telling Google exactly what a piece of content is. The moment the markup and the page disagree, the safest response for a search engine is to ignore it.

Does adding JSON-LD structured data improve search rankings?

Not directly - structured data isn't a ranking factor the way page speed or content quality are. What it does is make a page eligible for rich results like star ratings, FAQ accordions, and breadcrumb trails in search listings, which tend to improve click-through rate at whatever position the page already ranks. It only works if the markup exactly matches the visible page content.