Structured data: the five types actually worth adding

Structured data is a machine-readable description of what a page is about, embedded in the page. Instead of hoping a crawler works out that "£24.99" is a price and "4.6" is a rating, you state it explicitly in a defined vocabulary.

The vocabulary is schema.org and it has hundreds of types. Google produces rich results for roughly two dozen of them. Everything else is markup that validates perfectly and does nothing — which is worth knowing before you spend a week on it.

Use JSON-LD

Three formats exist. Microdata and RDFa interleave attributes through your HTML, which means every template change risks breaking the markup and vice versa. JSON-LD is a single script block in the head, entirely separate from the markup, and it is what Google explicitly recommends.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Structured data: the five types worth adding",
  "datePublished": "2026-09-03",
  "author": { "@type": "Organization", "name": "S2FreeTools" }
}
</script>

One block, generated server-side, independent of layout. There is no good reason to use the other two on a new site.

The types that earn their place

Organization or LocalBusiness — sitewide, on every page. Establishes the entity behind the site: name, logo, social profiles, and for a local business the address, opening hours and phone number. This is what feeds a knowledge panel, and for a business with a physical location it is the highest-value markup available.

BreadcrumbList — replaces the raw URL in a search result with a readable hierarchy: Home › Guides › Structured data. Cheap to add, applies to nearly every page, and measurably improves click-through on deep pages. Probably the best effort-to-benefit ratio in the list.

Article — for anything editorial. Headline, publication date, modified date, author, image. It feeds Top Stories and the date shown next to a result, and freshness is one of the few signals an AI answer engine has for preferring one page over an identical one.

Product with Offer and AggregateRating — for anything sold. Price, currency, availability and star rating appear directly in the result. On a commercial site this changes click-through more than anything else here. It also has the strictest rules: the price and rating must be genuinely visible on the page, and ratings must come from real users. Manufactured ratings are a manual-action risk, not a clever trick.

FAQPage — a caveat rather than a recommendation. Google dramatically restricted FAQ rich results in 2023; they now show almost exclusively for government and health sites. The markup is still worth adding because it helps AI answer engines extract question–answer pairs cleanly, but do not add it expecting the search feature. That feature is effectively gone for ordinary sites.

Similarly, HowTo rich results were retired outright. The markup validates; the visual result does not appear.

Rules that decide whether it works

The markup must describe what is visibly on the page. This is the single rule underneath all the others. Marking up a rating that is not shown, a price that differs from the displayed one, or FAQ answers that appear nowhere in the copy is a guidelines violation. Google is good at detecting it, and the penalty is losing rich results across the whole site — not just the page.

Required properties are required. Each type has a set Google needs, and one missing property means no rich result. Recommended properties are genuinely optional, though they improve what is shown.

Connect the nodes. Separate, unlinked blocks for Organization, WebSite and Article describe three unrelated things. Referencing them by @id — the article's publisher pointing at the organisation's node — describes one site. This is what a well-formed graph looks like, and it is a common gap.

Valid does not mean eligible. A schema validator checks the syntax against schema.org. Google's rich results test checks whether Google will do anything with it. Both are worth running; they answer different questions.

Rich results are never guaranteed. Correct markup makes a page eligible. Whether the feature shows depends on the query, the device and Google's judgement about the site. Correct markup that shows nothing is not necessarily broken markup.

Building and checking it

Our schema markup generator produces JSON-LD for the common types with the required properties present, and the schema validator checks what a live URL actually emits — which is frequently not what the CMS believes it emits, especially where a theme and a plugin both inject a block and produce two conflicting Organization nodes.

The part that is growing

Structured data has quietly become more valuable for a reason unrelated to search results. AI answer engines lean heavily on it, because unambiguous machine-readable facts are far easier to cite correctly than prose that has to be parsed and guessed at.

So the calculation has shifted. FAQ markup no longer earns a search feature, and it is still worth adding — not for the SERP, but because it makes the page's questions and answers legible to systems that increasingly decide whether your page is the one quoted.

Where to start

Sitewide Organization plus BreadcrumbList on every page, connected by @id. Then Article on editorial pages or Product on commercial ones. That covers the great majority of the available benefit. Everything past that is refinement, and a lot of it is markup nobody will ever read.