Schema markup for AEO is structured data, written in JSON-LD, that tells AI answer engines exactly what your content means so they can extract and cite it accurately. The schema types that matter most are Organization, Article or BlogPosting, FAQPage, Person, BreadcrumbList, HowTo and Speakable, bundled together with @graph and validated with the Google Rich Results Test. Schema does not guarantee citations, but without it AI engines have to guess, and they cite competitors who do not make them guess.

Why schema markup matters more for AEO than it did for SEO

For traditional SEO, schema was a helpful signal. For AEO, it is closer to a source. When a person searches Google, the engine ranks pages and the person clicks. When a person asks ChatGPT, Perplexity, Gemini or Google AI Overviews a question, the engine reads structured content, extracts the relevant facts, and writes an answer. Clean schema markup is how you hand those engines accurate, machine-readable facts instead of making them infer meaning from your HTML.

The shift is simple but important. Search engines could afford to treat schema as optional because they had decades of ranking signals to fall back on. AI answer engines are extraction machines: they reward content they can parse confidently and quietly skip content they cannot. A page with accurate Organization, Article and FAQPage schema is far easier to cite correctly than a visually identical page with none.

The schema types that matter for AEO

You do not need every schema type. You need the handful that describe who you are, what each page is, and which questions it answers. Here is the priority order for South African businesses.

Start with Organization, Article or BlogPosting, FAQPage and BreadcrumbList on every relevant page. Add Person, HowTo, Service and Speakable where they genuinely apply. Do not add schema for content that is not actually on the page.

Use JSON-LD, nothing else

Use JSON-LD for every piece of structured data and ignore Microdata and RDFa. JSON-LD sits in a single script block in your page head, keeps your markup separate from your content, and is the format Google, Bing and the major AI engines parse most reliably. Microdata and RDFa scatter attributes through your HTML, are harder to maintain, and offer no AEO advantage. Need a head start? Our free schema markup generator builds valid Organization, FAQPage, LocalBusiness and Article JSON-LD you can copy straight in.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://www.example.co.za/#organisation",
  "name": "Your Business",
  "url": "https://www.example.co.za/",
  "logo": "https://www.example.co.za/logo.png",
  "address": { "@type": "PostalAddress", "addressLocality": "Pretoria", "addressCountry": "ZA" },
  "sameAs": [
    "https://www.facebook.com/yourbusiness",
    "https://www.linkedin.com/company/yourbusiness"
  ]
}
</script>

The core schema types, explained

Organization schema, your foundation

Organization schema is the entity record for your business. It names you, locates you, and links to your verified profiles through the sameAs array, which is how AI engines confirm you are a real, consistent entity. Give every page the same Organization block, referenced by a stable @id.

{
  "@context": "https://schema.org",
  "@type": "Organization",
  "@id": "https://www.example.co.za/#organisation",
  "name": "Your Business",
  "url": "https://www.example.co.za/",
  "logo": "https://www.example.co.za/logo.png",
  "address": { "@type": "PostalAddress", "addressLocality": "Pretoria", "addressCountry": "ZA" },
  "sameAs": [
    "https://www.facebook.com/yourbusiness",
    "https://www.linkedin.com/company/yourbusiness"
  ]
}

Article and BlogPosting schema

Every editorial page should declare itself an Article or BlogPosting with headline, datePublished, dateModified, an author referenced by @id, and a publisher. Accurate dates matter: AI engines favour content that is demonstrably current, and a missing dateModified makes a fresh page look stale.

FAQPage schema, built for extraction

FAQPage schema is the most directly useful type for AEO because it hands engines clean question-and-answer pairs. Keep answers tight, roughly 40 to 60 words, factual, and self-contained so they can be quoted without surrounding context. Only mark up questions and answers that actually appear on the page.

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "How much does the service cost?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Packages start from R5,000 per month with no long-term contract."
    }
  }]
}

Person, BreadcrumbList and Speakable

Person schema, with a knowsAbout array, signals real author expertise, a genuine E-E-A-T input. BreadcrumbList tells engines where a page fits. Speakable marks the passages, usually your answer-first summary and key takeaways, that are safe to read aloud in voice and assistant results.

Bundle it with @graph and place it right

Put all of a page's schema in one JSON-LD block, connected with an @graph array and shared @id references. Bundling lets you link the Article to its author Person, its publisher Organization, and its breadcrumb without repeating data. It is cleaner for you to maintain and easier for engines to read as one connected entity graph than as several disconnected blocks.

Place the block in the page head. Keep one Organization definition site-wide and reference it everywhere by @id rather than redefining it. Never describe content in schema that a visitor cannot see on the page, that is the fastest way to get markup ignored or flagged.

Validate everything, it is not optional

Broken schema is worse than no schema. Before and after every change, run your markup through the Google Rich Results Test and the Schema.org Validator. Both are free, both catch the errors that quietly stop engines trusting your data, and both should pass clean before you publish.

Common schema mistakes that stop AI engines citing you

Most schema problems are not exotic. They are the same handful of mistakes repeated across thousands of South African sites.

• Phantom content: marking up FAQs, reviews or prices that are not visible on the page. Engines treat this as manipulation.
• Inconsistent @id: a different Organization identifier on each page, so engines never connect your pages into one entity.
• Missing sameAs: no links to your verified profiles, so your entity cannot be confirmed.
• Stale dates: a datePublished from years ago and no dateModified, making current content look abandoned.
• Wrong types: using Article on a product page or Service on a blog post.
• Unvalidated markup: a single syntax error that invalidates the whole block.
• Over-marking: stacking schema types that do not match the page just to look thorough.
• Microdata leftovers: old RDFa or Microdata fighting your JSON-LD and sending mixed signals.

“Schema does not make weak content rank. What it does is remove every excuse an AI engine has to misread, mistrust, or skip content that is genuinely good. Get the facts machine-readable and accurate, then let the content do the work.”

Cobus van der Westhuizen, Founder, Juicy Designs, reviewed and verified June 2026

One honest caveat on the newest layer: llms.txt. It is an emerging file that points AI crawlers at your most important pages, and it is worth adding as a low-cost foundation. But Google has said llms.txt is not a ranking signal, and no major AI engine has confirmed it as a citation factor. Schema markup is the machine-readable layer that actually does the work today. For the wider picture, see our guide to AI SEO, AEO and GEO and the difference between AEO and SEO.

Frequently asked questions