What structured data is
After this lesson you can write three common JSON-LD blocks and say honestly what they are for.
Structured data is a standard way to describe the content of a page to machines: this is an article, written by this person, published on this date, by this organisation. The shared vocabulary is schema.org. JSON-LD is the format Google recommends for it: a block of JSON inside a script tag, separate from the visible HTML.
What it does and does not do (documented)
- Google uses structured data to understand pages and to power certain rich results listed in its structured data gallery. As of 2026-11 that gallery includes, among others, Article, Breadcrumb, Local business, Organization, Product, Profile page, Q&A, Review snippet and Software app. (I read the gallery list myself; features come and go, as lesson FAQPage, HowTo and Q&A markup after the deprecations shows.)
- Google's AI features page says no special schema.org structured data is needed for AI Overviews and AI Mode, and its generative guide says over-focusing on structured data is unnecessary for those features. Both documented.
- Markup must represent visible content. Valid markup is not a promise of display.
So the honest purpose of structured data is: precise description, eligibility for certain rich results, and a consistent statement of facts. Not an AI shortcut.
Article
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Choosing a sourdough flour",
"datePublished": "2026-10-02",
"dateModified": "2026-11-14",
"author": {
"@type": "Person",
"name": "Asha Rao",
"url": "https://www.example.com/about/asha-rao/"
},
"publisher": {
"@type": "Organization",
"name": "Greenfield Bakery",
"url": "https://www.example.com/"
},
"image": "https://www.example.com/images/flour-test.jpg"
}All names, dates and addresses are fictional. Use real ones that match the page. Google's Article documentation treats most properties as recommended rather than required; check the current list.
Organization
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Greenfield Bakery",
"url": "https://www.example.com/",
"logo": "https://www.example.com/images/logo.png",
"sameAs": [
"https://www.example.org/greenfield-bakery-profile"
],
"contactPoint": {
"@type": "ContactPoint",
"contactType": "customer support",
"email": "hello@example.com"
}
}sameAs lists other pages that clearly represent the same organisation, such as an official profile. Only list profiles you control or that genuinely describe you.
Person
{
"@context": "https://schema.org",
"@type": "Person",
"name": "Asha Rao",
"jobTitle": "Head baker",
"worksFor": { "@type": "Organization", "name": "Greenfield Bakery" },
"url": "https://www.example.com/about/asha-rao/"
}Schema.org defines Person. Google's gallery has a Profile page feature for pages centred on a person or organisation, but I did not find a stand-alone Person rich result; treat Person markup as description and consistency (lesson Entities, brand mentions and consistency). Whether it affects AI answers is a hypothesis.
Putting it on a page
<script type="application/ld+json">
{ "@context": "https://schema.org", "@type": "Article", "headline": "Choosing a sourdough flour" }
</script>Check syntax first: paste your block into the JSON formatter. A trailing comma or a curly quote breaks the whole block silently. Then test in Google's Rich Results Test and check the Search Console enhancement reports, if available for your type.
Real example on this site
This site's lesson pages emit a LearningResource and a BreadcrumbList block, with an author of type Person and an isPartOf Course reference. Tool pages emit SoftwareApplication, BreadcrumbList and FAQPage. The audit script scripts/seo-audit.mjs validates the structured data of each exported page and checks that the breadcrumb data matches the visible breadcrumb, an example of the "match the page" rule enforced by machine. Which of these ever produce rich results is not known and not claimed.
Rules of thumb (established practice)
- Mark up only what is on the page.
- Keep dates and names identical to the visible text.
- One main entity per page where possible.
- Maintain it: stale markup is worse than none.
- Use official documentation for the properties of each type.
Common mistakes
- Marking up content hidden from users, or a rating nobody gave.
- Copying someone else's JSON-LD and leaving their name in.
- Using curly quotes or trailing commas.
- Letting
dateModifiedchange automatically on every build when nothing changed. - Believing markup replaces good visible content.
Exercise
- Write Article, Organization and Person JSON-LD for your own page using only true facts. Parse each in the JSON formatter.
- Compare each field to the visible text and list mismatches.
- Open Google's structured data gallery and note which of these types has a documented rich result, with the date you checked.
Show a sample answer
Step 3 result as of 2026-11: Article, Organization and Profile page appear in the gallery; Person does not have a separate entry. Record the date, because the list changes.
Previous: Trust, authorship and E-E-A-T. Next: FAQPage, HowTo and Q&A markup after the deprecations.
Key points
- Structured data describes page content in a standard vocabulary (schema.org); JSON-LD is the format Google recommends.
- Markup must match visible content; never mark up things users cannot see.
- Google's AI features page says no special schema is needed for those features; structured data is still useful for supported rich results.
- Article, Organization and Profile page are among the features in Google's structured data gallery as of 2026-11; Person has no dedicated rich result.
- Validate with a JSON parser and Google's Rich Results Test; valid markup does not guarantee any display.