Structured data explained: JSON-LD, rich results, and why it matters in 2026

The first time I added FAQPage schema to a support article, I made a small but common mistake: I marked up the visible FAQ accordion questions, but wrote slightly different, more keyword-stuffed answers in the JSON-LD than what actually appeared on the page. Google's Rich Results Test passed the markup as technically valid. Google's manual reviewers did not agree, and the rich result never appeared, because the structured data has to describe what a user genuinely sees, not an idealized version of it.

What structured data actually is, mechanically

Structured data is a standardized, machine-readable description of your page's content, written in a vocabulary called schema.org and most commonly embedded as a JSON-LD script block in the page head. Unlike your visible HTML, which crawlers have to parse and interpret probabilistically, JSON-LD hands the crawler an explicit, unambiguous statement: this is an Article, its headline is X, its author is Y, it was published on this date.

Article schema: the fields that actually unlock features

For a blog post, the Article or BlogPosting type needs headline, image, datePublished, dateModified, and author (as a Person or Organization object with a name) to be eligible for enhancements like the byline and date shown in some search result formats. Missing datePublished is one of the most common omissions I see, and it silently disqualifies otherwise-correct markup from date-related rich result features.

FAQPage schema and its 2023 restriction

As of an August 2023 policy change, Google now only shows the FAQ rich result for well-known, authoritative government and health websites — most other sites can still add valid FAQPage markup, but it typically won't produce the visible dropdown snippet it used to for everyone. It's still worth adding correctly since it may help with other forms of understanding and eligibility, but don't build an SEO strategy around FAQ rich results appearing for a general blog anymore.

Product schema and the review-snippet crackdown

Product structured data with aggregateRating and offers fields can produce star ratings and price display directly in search results, but Google has repeatedly tightened enforcement against self-serving or fabricated ratings; the aggregateRating value must reflect real, verifiable reviews collected on your own site, not review counts imported wholesale from elsewhere or fabricated to look impressive.

Validating markup the right way

Google's Rich Results Test (search.google.com/test/rich-results) checks against Google's specific subset of supported types and required/recommended fields, while the more general Schema.org Validator checks broader spec compliance across the full vocabulary regardless of whether Google currently uses that type. Use the Rich Results Test for anything you want to actually appear differently in Google search, and check Search Console's Enhancements reports periodically after deployment, since validation at test-time doesn't guarantee the rich result renders in live search indefinitely.

A minimal, correct JSON-LD example is worth more than a giant template

Rather than copying a 200-line JSON-LD template covering every possible optional field, start with the required fields for your specific type, validate, confirm it renders correctly, then add optional enhancements like breadcrumbs or video incrementally. Large templates copied wholesale from tutorials frequently include stale or incorrect fields that fail validation silently until you actually test.

Structured data FAQ

Does adding structured data improve my ranking position directly? No, structured data is not a direct ranking factor by Google's own statements, but it can improve click-through rate through richer result appearances, which is a real, measurable traffic gain even without a rank change.

Can I use Microdata or RDFa instead of JSON-LD? Yes technically, all three formats are supported, but Google explicitly recommends JSON-LD because it's easier to implement and validate without touching visible HTML markup, and it's what the overwhelming majority of modern sites use.

What happens if my structured data doesn't match the visible page content? It risks a manual action for structured data spam under Google's guidelines, and even without a manual action, mismatched markup commonly fails to produce any rich result at all.

Do I need a developer to add JSON-LD, or can I use a generator tool? A generator tool covers most common cases well for standard content types like articles or FAQs; custom or nested schema types for things like recipes with multiple sub-entities usually benefit from a developer's review before deployment.

Breadcrumb and sitelinks search box schema round out most sites

BreadcrumbList schema tells Google the hierarchical path to a page and commonly produces the breadcrumb trail shown under a result instead of the raw URL, which is a small but real trust and clarity improvement in the search listing. WebSite schema with a SearchAction can enable the sitelinks search box for brand-name searches of your own site, though eligibility for that specific feature depends on overall site authority, not just correct markup.

Testing structured data changes before they go live

Pushing untested JSON-LD straight to production risks silently breaking an existing rich result if a required field gets dropped during a template change, which can go unnoticed for weeks since nothing visibly breaks on the page itself. Running the Rich Results Test against a staging URL as part of your deployment checklist, the same way you'd run any other pre-launch check, catches this before it costs you real search visibility.

Nested schema for complex content types

A recipe page can nest NutritionInformation, AggregateRating, and a list of HowToStep objects all inside a single Recipe schema block, and getting the nesting structure exactly right is where most hand-written JSON-LD breaks down — a misplaced closing brace or wrong property name silently invalidates the outer object. For any schema type with more than two or three nested sub-objects, generating it programmatically from your actual content fields rather than hand-typing it avoids this entire category of error.

HowTo schema's own restriction mirrors the FAQ one

Similar to the 2023 FAQPage restriction, Google significantly narrowed HowTo rich result eligibility on desktop and reduced its presence generally, meaning the visible step-by-step rich result you might remember from a couple of years ago is now far less common even with fully valid markup. It's still reasonable to include for the semantic clarity it gives crawlers, but don't plan a redesign around HowTo rich results reliably appearing.