What is Schema Markup? The Complete Guide for 2026

Schema markup, defined

Schema markup is a small block of structured data you add to your web pages using a shared vocabulary called Schema.org. Its purpose is to translate the meaning of your content into a format machines can read with zero ambiguity. A human reading a page sees “$49.99” and instantly knows it's a price. A crawler sees a string of characters and has to guess. Schema markup removes the guessing.

Schema.org is jointly maintained by Google, Microsoft, Yahoo, and Yandex, with input from the broader web standards community. It defines hundreds of types — Person, Article, Product, Recipe, Event, Movie, Course — each with a specific set of properties. When you mark up a page as a Product, you can attach properties like name, price, availability, and aggregateRating — and every machine that reads your page understands them in exactly the same way.

The vocabulary is the standard. The format you embed it in is up to you. The three options — JSON-LD, Microdata, and RDFa — produce equivalent results, but they have very different practical trade-offs. The next section unpacks why JSON-LD has won.

JSON-LD vs Microdata vs RDFa

All three formats encode the same Schema.org vocabulary, but they live in very different parts of your HTML. Pick one and stick with it — mixing formats on the same page produces unpredictable results.

JSON-LD (recommended)

JSON-LD lives in a single <script type="application/ld+json"> block, usually placed in the page <head>. Because it's decoupled from the visible HTML, you can add, remove, or update structured data without touching your templates. CMS plugins can inject it. Static-site generators can build it. AI-search systems prefer it because the data is unambiguous and self-contained. Google explicitly recommends it.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "What is Schema Markup?",
  "author": { "@type": "Person", "name": "Jane Smith" },
  "datePublished": "2026-05-03"
}
</script>

Microdata (legacy)

Microdata weaves Schema.org attributes directly into your visible HTML using itemscope, itemtype, and itemprop. It worked well in 2014, but today it's fragile: a designer cleaning up the DOM can quietly delete an entire schema; a templating engine can break the structure with a stray wrapper. Use it only if you're maintaining a legacy site that already relies on it.

RDFa (specialty)

RDFa is conceptually similar to Microdata but uses a different attribute set (vocab, typeof, property) inherited from the broader Linked Data world. It's mostly used inside knowledge-graph and academic publishing systems. For a normal commercial site in 2026, JSON-LD is the correct default — there is no situation where RDFa or Microdata will outperform a clean JSON-LD implementation.

Why schema markup matters for AI search

For roughly twenty years, schema markup was a Google-rich-results play. You added FAQ schema to get an FAQ accordion in the SERP. You added Recipe schema to qualify for the recipe carousel. The benefit was visible, narrow, and traffic-driven.

Then large language models started indexing the web. ChatGPT's search, Perplexity, Google AI Overviews, Gemini, and Bing's Copilot all read the same structured data — but they use it for a fundamentally different purpose. Instead of rendering rich snippets, they extract entities, facts, and relationships and weave them into generated answers. Pages with clean schema become the cited sources behind those answers. Pages without it get summarized, paraphrased, or skipped entirely.

A 2025 industry study found that pages with complete Article and Organizationschema were cited by ChatGPT and Perplexity at roughly 4–6× the rate of equivalent pages without schema. The pattern isn't magic — it's mechanical. When the model needs to cite a source for “Who founded Acme Corp?”, it reaches for the page that explicitly declares an Organization with a named founder, not the page that mentions the founder in the third paragraph.

For AI search, the schema fields that matter most are the ones that ground your content in the AI's knowledge graph:

sameAs — links to your Wikipedia page, Wikidata entry, LinkedIn, GitHub, and social profiles. This is how AI systems verify you're the entity you claim to be.
author as Person — not just a name string. A nested Person object with a sameAs profile and a description gives the model an entity to attribute the content to.
datePublished and dateModified — AI engines weight freshness heavily. Stale dates are treated as a negative signal.
description — at 150+ words rather than the typical 30-word meta description. AI systems use this as a content summary when deciding what to extract.
mainEntity — explicitly declares the primary entity the page is about. Removes any ambiguity for the AI parser.

For a deeper breakdown, see our companion guide on schema markup for AI search, which covers the GEO (Generative Engine Optimization) framework end to end.

How to add schema markup to your website

Implementation is the part most guides over-complicate. The actual workflow has four steps:

1
Pick the right schema type
Match the schema type to the page's primary purpose. A blog post is an Article. A product page is a Product. A help page with Q&A is a FAQPage. You can combine multiple types on a single page (Article + FAQPage + BreadcrumbList is a common, valuable combo) — just don't pick the wrong primary type.
2
Generate the JSON-LD
Use a generator that outputs valid Schema.org JSON-LD. Our ten free generators cover the most common types and produce AI-optimized output by default — meaning every recommended field is included, not just the bare minimum Google requires.
3
Embed it on the page
Paste the <script type="application/ld+json"> block into your page's <head>. WordPress users can use a plugin like Rank Math, WPCode, or Yoast. Shopify users paste into the relevant theme template. Webflow, Wix, Squarespace, and most page builders have a custom-code field.
4
Validate before shipping
Run the page through our schema validator (catches AI-readiness issues) and Google's Rich Results Test (confirms rich-result eligibility). Both checks take 30 seconds and catch ~95% of real-world mistakes.

That's it. The only step that takes meaningful effort is the first one — picking the right type — and even that becomes second nature after you've marked up a few pages. Everything downstream is mechanical.

The 10 schema types most websites need

Schema.org defines hundreds of types, but a normal commercial website only ever uses a handful. Here are the ten that cover roughly 95% of real-world schema needs, each linked to a free generator that outputs AI-optimized JSON-LD.

FAQ Schema

Mark up question-and-answer content. The single highest-leverage schema for AI citation.

Article Schema

Identify long-form content — blog posts, news, tutorials — with author and publish dates.

Product Schema

Pricing, ratings, availability, and SKUs for ecommerce listings and rich product cards.

Local Business

Address, hours, and phone for brick-and-mortar stores, restaurants, and service businesses.

HowTo Schema

Step-by-step instructional content. Critical for voice search and AI step extraction.

Organization

Site-wide brand entity with sameAs links to social profiles and Wikipedia/Wikidata.

Review Schema

Editorial reviews and aggregated ratings for products, services, and local businesses.

Event Schema

Conferences, concerts, webinars, and classes with dates, venue, and ticketing info.

Breadcrumb

Signal site hierarchy on deep pages. Helps both users and AI map your information architecture.

Video Schema

Mark up videos with thumbnail, duration, upload date, and transcript URL for video AI surfaces.

Common schema mistakes (and how to avoid them)

After validating thousands of schema blocks through our tools, the same handful of mistakes account for the vast majority of failures. None are hard to fix once you know to look for them.

Schema that doesn't match visible content. Adding FAQ schema for questions that don't actually appear on the page violates Google's guidelines and can trigger a manual penalty. Schema must mirror what a human reader sees.
Author as a string, not a Person. "author": "Jane Smith" is technically valid but tells the AI nothing. Use a nested Person object with name, url, and sameAs at minimum.
Missing or stale dateModified. AI engines treat freshness as a major weighting signal. A page with datePublished: "2019-03-12" and no dateModified reads as abandoned, even if the content is current.
Relative URLs in image and url fields. Use absolute URLs (https://example.com/img.png), never relative paths. Crawlers and AI parsers don't always resolve relative URLs the way browsers do.
Duplicating schema across the site. If your homepage and your /about page both declare the same Organization schema, that's fine. But emitting two identical Article schemas on the same page (one from your CMS, one from a plugin) creates conflict. Audit your raw HTML before going live.
Skipping the validator. JSON syntax errors silently break the entire schema block. A missing comma or unescaped quote means the schema isn't parsed at all. Always validate before publishing.

Frequently asked questions

What is schema markup in simple terms?+

Schema markup is structured data you add to your web pages so search engines and AI systems can understand what the page is actually about. Instead of guessing whether a number on the page is a price, a phone number, or a year, schema markup explicitly labels it. It's a vocabulary defined by Schema.org and used by Google, Bing, ChatGPT, Perplexity, Gemini, and every other major search and AI engine. The most common format is JSON-LD, a small JavaScript-Object-Notation block placed in the page's HTML head.

Is schema markup the same as SEO?+

Schema markup is one component of modern SEO, but it isn't the same thing. SEO is the broader discipline of optimizing pages for search visibility, covering content quality, technical performance, internal linking, backlinks, and user experience. Schema markup is the specific technical layer that tells machines what your content means. You can rank on traditional Google with no schema at all, but you'll struggle to be cited by AI search engines without it. In 2026, schema is closer to mandatory than optional for any site that wants AI visibility.

Does schema markup help with Google rankings?+

Schema markup is not a direct ranking signal in Google's traditional blue-link results — Google has stated this clearly. What it does is unlock rich results (FAQ accordions, star ratings, recipe cards, event listings, product prices) that increase click-through rates from the search results page. It also feeds Google's AI Overviews, which now decide whether your page is cited above the traditional results. So while schema doesn't raise your position, it dramatically improves how your listing performs once it's there.

What's the difference between JSON-LD, Microdata, and RDFa?+

All three are formats for embedding Schema.org structured data in HTML. JSON-LD places the data in a separate <script type="application/ld+json"> block, decoupled from the visible HTML. Microdata weaves attributes (itemscope, itemtype, itemprop) directly into HTML elements. RDFa is similar to Microdata but uses different attribute names. Google explicitly recommends JSON-LD because it's easier to maintain, doesn't break under HTML edits, and lets you put all your structured data in one place. For new implementations in 2026, use JSON-LD.

Where do I put the schema markup code on my page?+

Schema.org and Google recommend placing JSON-LD inside a <script type="application/ld+json">…</script> block in the <head> section of your HTML. It also works correctly in the <body>, so don't worry if a CMS forces it elsewhere. For WordPress, plugins like Rank Math or WPCode insert it for you. For Next.js, use the next/script component or a static <script> tag rendered server-side. The important rule is exactly one schema block per page section — don't duplicate identical schema in multiple places.

How do I know if my schema markup is working?+

Use two tools. First, our free schema validator catches structural errors and AI-readiness issues before you publish. Second, Google's Rich Results Test confirms whether your schema is eligible for Google's rich results. After publishing, monitor Google Search Console's Enhancements section for FAQ, Article, Product, or other rich-result reports. For AI engines, the only real test is asking Perplexity, ChatGPT, or Gemini a question your page should answer and seeing whether it cites you.

Do I need schema markup if I have great content?+

Yes. Great content makes you worth citing; schema markup makes you machine-readable enough to be cited reliably. AI engines parse hundreds of millions of pages a day and rely on structured signals to identify high-quality, well-described content quickly. A page with excellent content and no schema is at a disadvantage versus a page with similar content and complete schema. The two work together — neither replaces the other.

How long does schema markup take to start working?+

For Google, expect 1–4 weeks before you see rich results in search. The page must first be re-crawled and the schema validated. For AI search engines (ChatGPT, Perplexity, Google AI Overviews), the timeline is less predictable — citation patterns can shift within days as models re-index, or take longer for newer pages without established authority. Once schema is in place, monitor weekly for the first month, then monthly thereafter.

Ready to add schema to your site?

Pick a schema type, fill a quick form, copy the JSON-LD. No sign-up, no paywall, no usage limits — every generator outputs AI-optimized markup by default.

Browse all 10 generators Read the AI-search guide