What structured data is and what it does
Search engines read HTML well enough to index text, but a page's meaning — which number is the price, which name is the author, whether the stars are a rating — is ambiguous to a machine. Structured data (Schema MarkupStructured data vocabulary.) resolves the ambiguity with a vocabulary (schema.org, maintained jointly by the major engines) and a format (JSON-LD, which Google recommends).
Google documents two uses. Rich results: for specific types, the listing on the SERPThe page of results a search engine shows. can show extra elements — product price and availability, review stars, recipe times, event dates, breadcrumbsA trail of links showing a page's position in the site hierarchy, from the home page down to the current page., sitelinks search boxes, video details — which change click behaviour on that listing. Understanding: markup helps the engine identify entities (an organisation, a person, a product) and connect pages to them, which supports knowledge panels and disambiguation. What Google does not document is any general ranking benefit from markup, and claims that "schema improves rankings" are inference at best.
The types that matter
Google's search gallery lists the supported types and their required and recommended properties; the list changes, and the gallery is the reference. Types that most sites should consider:
| Page type | Schema type | What it can show | Notes |
|---|---|---|---|
| Any page with a hierarchy | BreadcrumbList | The path in the result instead of the URL | Cheap and broadly useful; this site emits it on every page |
| Organisation homepage | Organization (with logo, contact, social profiles) | Knowledge panel inputs | Entity identity; one per site |
| Person or author page | Person | Entity identity; author attribution | Used with Article author |
| Articles and guides | Article (or NewsArticle, BlogPosting) | Headline, date, author in some surfaces | Dates must be real; see this site's own usage |
| Product pages | Product with Offer | Price, availability, rating, shipping | The highest-value commercial type; see category and product pages |
| Local business | LocalBusiness (and subtypes) | Address, hours, phone | Complements, does not replace, the Business Profile (local SEO) |
| Events | Event | Dates, location, tickets | Strict rules on what counts as an event |
| Recipes, videos, job postings, courses | Their respective types | Rich cards | Each has detailed requirements |
| Reviews | Review / AggregateRating | Stars | Self-serving reviews on your own organisation are not eligible; review markup must reflect real, visible reviews |
| FAQ sections | FAQPage | Expanded questions | Rich result restricted to government and health sites since 2023; still valid markup |
| Step-by-step instructions | HowTo | (none) | Rich result removed in 2023; this site does not use it |
Two facts from Google's policies shape the list: markup must describe content visible on the page, and must not be used to misrepresent the page (fake reviews, prices not shown, FAQ content hidden from users). Violations can produce a manual action that removes all rich results for the site. Two tutorials apply these rules to the commonest commercial types: product structured data for stores and LocalBusiness structured data for location pages.
Implementing JSON-LD
JSON-LD is a script block of type application/ld+json containing an object with @context set to
https://schema.org, an @type, and the properties the type requires. Properties that matter in practice:
@idgives an entity a stable identifier (for examplehttps://example.com/#organization) so that references from other pages' markup point to the same thing. Author on an article references the person entity by@idrather than restating it.- Nesting versus referencing. An
Offeris nested in itsProduct; an authorPersonis referenced from manyArticlenodes. Nest what belongs to one page; reference what is shared. - One graph per page. Several nodes in one script (or an
@grapharray) are fine; contradictory nodes — twoProductnodes with different prices — are not. - Match the page. The product's price in the markup is the price on the page; the article's headline is the visible headline; the breadcrumb matches the visible trail.
- Generate from data, not by hand. Markup written per page drifts; markup generated from the same data
that renders the page stays consistent. This site's content renderer emits
Article,CollectionPage,BreadcrumbListandFAQPagefrom frontmatter for exactly this reason. Where an AI tool drafts the markup, the tutorial on checking AI-generated structured data asks it for a field mapping the template can use, not finished markup for one page.
A product page's graph
A product page carries one script with three connected nodes: a Product (name, image, description, SKU,
brand reference, and an Offer with price, currency, availability and a URL that matches the canonical), an
AggregateRating nested in the product reflecting the visible reviews, and a BreadcrumbList for the
category path. The organisation node lives on the homepage with an @id, and the product's brand references
it. When the price changes in the catalogue, the markup changes with it because both render from the same
record. When the product goes out of stock, availability changes to the documented out-of-stock value rather
than the markup being removed.
Validation and monitoring
Three tools, three jobs:
- The schema.org validator checks that markup is syntactically valid and uses real types and properties.
- Google's Rich Results Test checks whether a page's markup is eligible for Google rich results and reports missing required or recommended properties for the supported types. Eligibility is not a guarantee of display.
- Search Console's enhancement reports (GSCGoogle's search performance and indexing tool.) show, per type, how many URLs are valid, have warnings or have errors across the site, and the search appearance filter in the performance report shows how rich results perform. These are the monitoring layer: a template change that breaks markup shows up here as errors within days.
The SEO page checker reports whether a live page carries structured data and which types, as a quick first check.
Deciding what to implement
Not every type is worth the effort. A short decision sequence:
- Is there a documented rich result for this page type, currently supported on sites like ours? If not, markup for understanding (Organization, Person, Breadcrumb, Article) is still worthwhile; markup chasing a rich result that will not show is not.
- Does the page genuinely contain what the type describes, visibly? Reviews that exist, prices that are shown, events that happen.
- Can the markup be generated from data? If it would be hand-written per page, plan the data source first.
- What is the expected effect? Product and review rich results change click-through (CTRShare of impressions that result in a click.) on commercial listings measurably; breadcrumbs tidy the listing; organisation markup supports brand identity. Prioritise by traffic to the template and by what the result adds.
Keeping markup correct over time
Markup rots quietly: a template change drops a property, a product goes out of stock and the markup still says in stock, a review widget changes its output, a new page type launches without any. The maintenance: generate markup from the same data the page displays, never from a parallel source; include the markup in the template tests so a change that breaks it fails the build; watch the enhancement reports in Search Console (GSCGoogle's search performance and indexing tool.) weekly for new errors by type; re-test one page per template after every release with the rich results test; and review once a year which types Google still supports, since it retires rich result types from time to time and markup for a retired type is wasted effort.
Reading the enhancement reports
Each supported type has a report listing valid items, items with warnings and items with errors, with the specific property at fault. Errors mean the item is ineligible for the rich result; warnings mean a recommended property is missing. The common findings: a required field empty because the source data was empty (fix the data or omit the item), a value in the wrong format (dates, prices, durations), a rating without a count, an image that is too small or not crawlable, and markup for content not on the page (remove it; it is a policy issue). Validating a fix in the report triggers a recheck of the affected URLs, which takes days.
Related markup that is not schema.org
Two other kinds of machine-readable annotation are often confused with structured data: Open Graph and Twitter card tags (which control how pages look when shared on social platforms, not in search) and the canonical link (Canonical TagTells search engines the preferred URL.), which is an indexing signal. All three live in the page head; only schema.org markup drives rich results. The title tags guide covers the other head elements.
When valid markup still produces no enhancement in results, work through the diagnosis on rich results not showing in Google.
Common mistakes
- Marking up what is not on the page. Hidden FAQ content, prices not shown, reviews not displayed — policy violations with site-wide consequences.
- Self-reviews. Rating markup on an organisation's own pages about itself is ineligible.
- Expecting rankings. Markup changes how a listing looks and how the engine understands the page, not where it ranks.
- Hand-written markup that drifts. The price in the markup is last year's.
- Chasing removed rich results. HowTo and (for most sites) FAQ markup show nothing; the effort belongs elsewhere.
- Valid but useless. Markup that passes validation on a type nobody searches for.
- No monitoring. Errors from a template change sit unnoticed until the rich results disappear.
The crawling and indexing guide explains how markup is read as part of indexing; the e-commerce section covers product markup in depth; and the Google Business Profile tutorial explains the relationship between local business markup and the profile that actually drives local results.
