Schema Markup in 2026: The Complete Guide to Structured Data for Google and AI Search

What it actually does, what changed after the March 2026 core update, and the exact code for the schema types that matter most

Schema markup spent its first decade doing one job: telling Google enough about a page that Google would reward it with a star rating, an FAQ accordion, or a recipe card in the search results. That job still exists in 2026, though it has narrowed considerably. But schema has quietly picked up a second, more consequential job over the past eighteen months — helping AI systems like Google's AI Overviews, ChatGPT, and Perplexity decide whether your business is a credible, citable source when they answer a question a potential customer asked them directly, with no blue links involved at all.

71% — of pages cited by ChatGPT include structured data, compared to a much lower baseline among uncited pages (industry analysis, 2026)

~65% — of pages cited in Google AI Mode answers include structured data (industry analysis, 2026)

3.2× — more likely a page with FAQPage schema is to appear in Google AI Overviews compared to a page without it (Frase.io testing, 2026)

This guide covers what schema markup actually is, what genuinely changed after Google's March 2026 core update narrowed eligibility for several widely-abused schema types, the five schema types that matter most for any small business website in 2026, and — because hype has outpaced honesty in much of the content written about this topic — an honest reality check on what schema can and cannot do for your AI visibility. Working code examples are included throughout, ready to adapt for your own site.

1. What Schema Markup Actually Is

Schema markup is structured data — a standardised way of describing the content of a web page in a format machines can read directly, rather than having to infer meaning from ordinary sentences. Without schema, a search engine reading a page about a plumbing business has to guess, from context and natural language processing, that "Marshall Plumbing" is a business name, "07700 900000" is a phone number, and "4.8 out of 67 reviews" is an aggregate rating. With schema, all three facts are explicitly labelled in a standard vocabulary (schema.org) that every major search engine and AI crawler recognises identically.

📄 PRIMARY SOURCE — Google's own format recommendation

Google's official Search Central documentation states that Google recommends JSON-LD for structured data because it is the easiest format to implement and maintain at scale. This recommendation has held consistently since 2017 and remains the current 2026 guidance. JSON-LD (JavaScript Object Notation for Linked Data) is a single, self-contained block of code placed in a page's head section — it does not require wrapping visible page content in special tags the way the older Microdata format did, which makes it far easier to implement correctly and far harder to break accidentally during a redesign.

Bing, DuckDuckGo, and the crawlers used by major AI systems — including GPTBot (OpenAI), ClaudeBot (Anthropic), and PerplexityBot — all parse JSON-LD reliably. Older formats like Microdata and RDFa still technically work but are increasingly fragile and unnecessary; there is no reason for a website built or rebuilt in 2026 to use anything other than JSON-LD.

2. What the March 2026 Core Update Actually Changed

Google's March 2026 core update, completed on March 12, produced the most significant shift in structured data strategy since rich snippets first launched — and the change is more nuanced, and more honest, than most coverage of it suggested at the time.

2.1 Rich Result Eligibility Narrowed for Abused Schema Types

FAQ, Review, and How-To schema had all been widely abused in the years leading up to 2026 — websites adding FAQPage markup to pages where the "questions" were thinly disguised keyword targets rather than genuine reader questions, or adding Review schema to editorial comparison posts that were not genuine product reviews at all, purely to win the visual real estate of a star rating in the search results. The March 2026 update specifically targeted this pattern. FAQ rich result impressions dropped by nearly half across tracked sites industry-wide. How-To rich results disappeared almost entirely from pages where the markup described supplementary content rather than the page's primary purpose. Review schema on editorial comparison content was demoted or, in cases of clear manipulation, actioned directly.

2.2 Genuine, Intent-Matched Schema Was Unaffected — Or Improved

Sites that had implemented schema aligned to genuine content intent — not as a search-result manipulation tactic — retained their rich result eligibility and, in many tracked cases, saw improved performance as the update demoted the manipulative competition around them. A page with a genuine, business-specific FAQ section answering real customer questions, marked up honestly with FAQPage schema, was not the target of this update. A page that artificially generated ten generic "frequently asked questions" purely to claim the visual space of an FAQ accordion was.

2.3 The Quieter, More Important Shift: Schema as a Trust Signal for AI Mode

Largely overlooked in coverage of the rich-result changes was a parallel shift: Google's Gemini-powered AI Mode now uses schema markup to verify claims, establish entity relationships, and assess source credibility during answer synthesis — independently of whether that schema produces any visible rich result at all. This is the single most important strategic change in how structured data should be thought of in 2026: schema that accurately describes your content increases the probability of AI Mode citation even when no traditional star-rating or FAQ box ever appears in the search results for that page.

⚖️ REALITY CHECK — The honest counterpoint — schema is not a silver bullet for AI citation

Search Engine Land's analysis, drawn from direct conversations with Google and Microsoft representatives, is appropriately cautious: structured data gives an advantage, but LLM systems prioritise relevance, topical authority, and semantic clarity over the simple presence or absence of markup. Adding schema to a thin or irrelevant page will not make AI systems cite it. Schema is connective tissue that helps machines verify genuinely good content faster — it is not a substitute for that content being genuinely good.

This is consistent with everything covered in our E-E-A-T guide: schema strengthens existing trust and expertise signals by making them machine-readable. It does not manufacture trust or expertise that is not already present in the content itself.

3. The Five Schema Types That Matter Most for a Small Business in 2026

Of the hundreds of schema types defined at schema.org, five cover the overwhelming majority of what a service business, local business, or content-publishing website needs in 2026.

3.1 Organization / LocalBusiness Schema

This is the foundational schema every business website should have on its homepage, establishing the core facts about the business: name, address, phone number, opening hours, service area, and — critically for citation purposes — links to the business's other verified profiles (Google Business Profile, LinkedIn, Trustpilot) via the sameAs property, which helps AI systems and search engines confirm that all these profiles refer to the same real-world entity.

{
"@context": "https://schema.org",
"@type": "LocalBusiness",
"name": "Marshall Brickwork & Construction",
"image": "https://example.com/logo.jpg",
"telephone": "+44 1634 000000",
"address": {
"@type": "PostalAddress",
"streetAddress": "12 High Street",
"addressLocality": "Rochester",
"addressRegion": "Kent",
"postalCode": "ME1 1AA",
"addressCountry": "GB"
},
"areaServed": ["Rochester", "Maidstone", "Sittingbourne", "Chatham"],
"openingHours": "Mo-Fr 08:00-18:00",
"aggregateRating": {
"@type": "AggregateRating",
"ratingValue": "4.9",
"reviewCount": "67"
},
"sameAs": [
"https://www.google.com/maps/place/...",
"https://www.linkedin.com/company/..."
]
}

3.2 FAQ Page Schema

FAQPage schema marks up genuine question-and-answer content in a format AI systems and search engines can extract directly. As covered in Section 2.1, this schema is now scrutinised more closely for genuine intent — but a real FAQ section, addressing real customer questions in your own words, remains one of the highest-value schema implementations available, with industry testing showing pages using it are over three times more likely to appear in Google AI Overviews.

{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [{
"@type": "Question",
"name": "Do you cover emergency call-outs in Maidstone?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Yes. We offer emergency call-out cover across Maidstone, Sittingbourne and the wider Medway area, with response within the hour during business hours."
}
}]
}

3.3 Article / BlogPosting Schema

For every blog post or guide on a website — including each article in this very series — Article schema establishes the headline, author, publication date, and last-updated date in a machine-readable format. The dates matter more than they might appear: as covered in our E-E-A-T guide, content freshness is a confirmed trust signal, and a clearly marked dateModified field is one of the clearest ways to demonstrate that a page has been kept current.

{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Why Your Website Fails Google PageSpeed",
"author": {
"@type": "Person",
"name": "Vasi",
"sameAs": "https://www.linkedin.com/in/..."
},
"publisher": {
"@type": "Organization",
"name": "WebWise Digital"
},
"datePublished": "2026-04-12",
"dateModified": "2026-06-10"
}

💡 TIP — Person schema with sameAs is the most overlooked detail

Industry data from SearchEngineLand (February 2026) found that articles with well-nested Person, Organization, and Article schema — where the Person schema includes a sameAs link to a real LinkedIn profile — saw AI Overview citations increase up to threefold compared to Article schema alone with no linked author entity. Without a verifiable Person entity, AI systems cannot reliably attribute expertise to a named individual, which directly undermines the Experience and Expertise signals covered in Section 5 of our E-E-A-T guide.

3.4 Service Schema

For any business offering distinct services, Service schema describes each one individually — far more precise than describing all services in a single paragraph of body text. This connects directly to the page-structure advice in our CRO guide: one page, one service, one Service schema block, matched precisely to the specific search intent that page targets.

{
"@context": "https://schema.org",
"@type": "Service",
"serviceType": "Emergency Boiler Repair",
"provider": {
"@type": "LocalBusiness",
"name": "Marshall Brickwork & Construction"
},
"areaServed": "Rochester, Kent",
"description": "24/7 emergency boiler repair across Rochester and the Medway towns, with same-day response for no-heat call-outs."
}

3.5 Review / AggregateRating Schema

Review schema, used honestly, marks up genuine customer reviews and the resulting aggregate star rating. As covered in Section 2.1, this is one of the types Google scrutinised most heavily in the March 2026 update — it must reflect real, verifiable reviews (ideally sourced and synced from Google Business Profile, as covered in our GBP guide) and must not be applied to content, like editorial comparisons of competitors' products, where no genuine first-party review relationship exists.

4. Implementing JSON-LD Schema: The Practical Steps

4.1 Where the Code Goes

A JSON-LD schema block sits inside a <script type="application/ld+json"> tag, placed in the page's <head> section. On a hand-coded site built with a modern framework like Next.js, this is typically generated dynamically per page — the homepage gets LocalBusiness schema, each blog post gets Article schema, each service page gets Service schema — rather than a single static block applied uniformly across the whole site.

4.2 The Most Common Implementation Mistakes

• Schema that does not match visible page content: If your FAQPage schema lists five questions but only two appear as visible text on the page, this mismatch can be flagged by Google as deceptive structured data, risking the schema being ignored or, in repeated cases, a manual action against the site.

• Dynamic schema that never renders: On sites using heavy client-side JavaScript to inject schema after the page loads, crawlers that do not execute that JavaScript fully may never see the schema at all. The reliable verification method, covered in Section 4.3 below, exists specifically to catch this.

• Disconnected entities: Article, Person, and Organization schema declared as three separate, unlinked blocks rather than properly nested together using @id references. As covered in Section 3.3, properly nested entity relationships are what allow AI systems to confidently attribute expertise and credibility — disconnected schema blocks lose much of this benefit even though each individual block is technically valid.

• Missing or incorrect required fields: Each schema type has fields Google considers required for rich result eligibility. Omitting them does not break the page, but it does disqualify that schema block from contributing to any visible rich result.

4.3 Verifying What Crawlers and AI Systems Actually See

The most reliable way to confirm that schema is genuinely present in what a crawler receives — not just what a human sees rendered in a browser — is to fetch the raw page source the way a bot would, before any client-side JavaScript executes:

curl -A "Mozilla/5.0" https://yourdomain.com/your-page/ \
| grep -A 200 'application/ld+json'

If the JSON-LD block appears in this output, crawlers and AI systems will see it. If it does not appear — because it was injected dynamically by JavaScript after the initial page load — it is effectively invisible to most crawlers, regardless of how correctly it was written. This single check catches one of the most common and least visible schema failures on JavaScript-heavy websites.

For validation of the schema's correctness once confirmed present, Google's own Rich Results Test and the Schema.org Validator both check syntax and required fields, flagging errors before they reach production.

5. Schema and AI Search: Setting Realistic Expectations

Given the volume of hyperbolic content published about schema and AI visibility in the past year — claims of "55% visibility boosts" and similar figures from sources with clear commercial incentive to oversell the tactic — it is worth stating plainly what the more rigorous, cross-referenced evidence actually supports.

The practical synthesis: implement schema properly because it strengthens signals that already matter (entity clarity, content freshness, trust verification) and because the downside risk is close to zero when done honestly. Do not implement schema as a substitute for the underlying work — genuine expertise, genuine experience, genuine authority — covered throughout this entire blog series. Schema is connective tissue, not content.

6. Schema by Business Type: What to Prioritise

7. The 15 Keywords This Article Targets

Conclusion: Build the Substance First, Then Make It Legible to Machines

Schema markup's reputation has swung from "nice-to-have technical detail" to "the secret to AI search visibility" in the space of about two years, and neither extreme is accurate. What the evidence — Google and Microsoft's own statements, the March 2026 core update's narrowing of abused rich result types, and the replicated testing on FAQPage citation rates — actually supports is more modest and more durable: schema is the format that lets machines verify, quickly and reliably, facts about your business that are already true and already present in your content. It does not invent expertise, experience, authority, or trust that is not otherwise there.

The five schema types in Section 3, implemented honestly and verified using the curl check in Section 4.3, cover the overwhelming majority of what any small business website needs in 2026 — whether the goal is a star rating in a traditional Google search result or a citation in an AI-generated answer. Every WebWise website build includes this schema foundation by default, nested correctly, and verified against live AI crawlers before launch.

If you would like your current site audited for what schema is actually present versus what AI crawlers can actually see, the starting point is a 15-minute call at webwise.digital/contact.