71% of pages that ChatGPT cites use schema markup. The pages it skips mostly don't.
That's the easy part to understand. Here's the uncomfortable part: adding schema wrong actually performs worse than not having it at all. Minimally-populated schema earns a 41.6% AI citation rate. No schema? 59.8%. You can implement structured data and make your AI visibility worse.
Most people don't know this because most guides stop at "here are the schema types, add them." This one doesn't. We'll cover which types move the needle in AI search, how to implement them in JSON-LD with templates you can actually use, and - crucially - what "implementing it right" actually means if you want it to do anything for citations. If you're familiar with answer engine optimization and want to get the technical layer right, keep reading.
Why Structured Data Works Differently in AI Search
AI engines don't rank pages. They synthesize answers - pulling from multiple sources, weighing credibility signals, and deciding what to extract. Structured data is one of those signals.
When Google's Gemini-powered AI Mode hits your page, it uses schema to verify your claims, map entity relationships, and assess whether your content is worth citing. Not to display a rich result. To decide if it trusts you.
Why JSON-LD and Not the Other Formats
JSON-LD lives in a <script> block in your page's <head>, completely separate from your HTML. AI crawlers parse it independently of your page layout - so even a messy, poorly-structured page can still feed clean structured signals to AI systems if the JSON-LD is correct.
Microdata and RDFa are technically valid, but they're embedded inside your HTML markup. They're harder for AI systems to parse cleanly, and they're not where the ecosystem is heading. JSON-LD is the standard. Use that.
The Gap Is Big Enough to Care About
65% of pages cited in Google AI Mode use schema markup. Pages with structured data get cited 3.2 times more often than those without.
Worth noting: that 3.2x figure is for well-implemented schema, not schema in general. The average includes all the half-finished implementations dragging the number down. Done properly, the impact is likely higher.
Which Schema Types Actually Get You Cited?
| Schema Type | AI Citation Impact | Priority |
|---|---|---|
| FAQPage | 2.7x higher citation rate | Tier 1 |
| Article | Baseline - establishes content identity | Tier 1 |
| Organization | Entity recognition across AI knowledge graphs | Tier 1 |
| HowTo | High for instructional / step-by-step content | Tier 2 |
| Product + AggregateRating | Important for product and review pages | Tier 2 |
| BreadcrumbList | Signals site architecture and topical context | Tier 2 |
Start with FAQPage - It's Not Close
FAQPage schema achieves citation rates 2.7 times higher than pages without it. The reason is almost obvious once you see it: FAQPage mirrors exactly how AI engines present answers - question, then answer. When your content is already structured that way, the AI doesn't have to interpret your intent. It just extracts.
Add Article schema to every post. It tells AI engines the content type, publication date, author, and publisher. Without it, an AI crawler has to guess all of that from context. Sometimes it guesses right. Often it doesn't.
Organization schema is the one most B2B SaaS brands skip, and it's a mistake. It's what connects your brand name, website, social profiles, and description into a coherent entity that AI knowledge graphs can recognize. If you want ChatGPT to know that your company exists and operates in a specific category - not just that a page on your domain has content - Organization schema is how that happens. There's more on the brand authority side of this in the guide to getting mentioned by ChatGPT.
Tier 2: Useful, But Not Your First Priority
HowTo schema is worth adding to any step-by-step content. Product and AggregateRating matter if you have product pages or review sections. BreadcrumbList helps AI systems understand your topical hierarchy.
These don't move citation rates as dramatically as the Tier 1 types on their own, but about 61% of pages cited in AI responses use three or more schema types - so they compound. Get the Tier 1 types right first, then layer these in.
The Part That Surprised Us: Generic Schema Hurts
We mentioned this at the top, but it's worth sitting with for a second.
Attribute-rich, fully-populated schema earns a 61.7% AI citation rate. Generic, minimally-populated schema - the kind where you've added the type but left most fields empty - earns 41.6%. No schema at all: 59.8%.
So if you've added schema to your site but left half the fields blank, you may have made things worse. An incomplete schema creates ambiguity. It signals that someone started to describe this content but didn't finish. AI systems are trying to assess confidence, and a partial signal doesn't build confidence - it undermines it.
If you implement it, do it completely. Every field.
How to Implement JSON-LD (Templates Ready to Copy)
Article Schema
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Your Post Title Here",
"description": "Your meta description here",
"author": {
"@type": "Organization",
"name": "Your Company Name",
"url": "https://yoursite.com"
},
"publisher": {
"@type": "Organization",
"name": "Your Company Name",
"logo": {
"@type": "ImageObject",
"url": "https://yoursite.com/logo.png"
}
},
"datePublished": "2026-04-09",
"dateModified": "2026-04-09",
"mainEntityOfPage": {
"@type": "WebPage",
"@id": "https://yoursite.com/blog/your-post-slug"
}
}
One thing people get wrong: they set dateModified equal to datePublished and never update it. Keep that field current when you revise the post. Freshness matters - content updated in the past three months averages nearly double the citation rate of stale pages.
FAQPage Schema
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Your question exactly as written in the heading?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Your complete, self-contained answer. Don't truncate it or reference another section. Write the full answer you'd want an AI to surface on its own."
}
},
{
"@type": "Question",
"name": "Second question?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Complete answer for question two."
}
}
]
}
The "name" field should match your heading text exactly. The "text" in the answer needs to be complete - not "see the section above." AI engines extract these directly and use them as standalone answers. Write them that way.
Organization Schema
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Your Company Name",
"url": "https://yoursite.com",
"logo": "https://yoursite.com/logo.png",
"description": "One-sentence description of what your company does.",
"sameAs": [
"https://twitter.com/yourhandle",
"https://linkedin.com/company/yourcompany",
"https://www.g2.com/products/yourproduct"
]
}
Most people skip the sameAs array. Don't. Those links connect your organization entity to your presence across the web - social profiles, G2, Capterra, product directories. AI knowledge graphs use them to confirm that your brand's identity is consistent across independent sources. Add every platform where your brand legitimately shows up.
Does Good Schema Guarantee Citations?
Honestly, no - and it's worth understanding why before you spend a weekend implementing everything above.
Schema tells AI engines what your content is and who created it. It doesn't tell them whether your content is worth recommending. That's still determined by content quality, topical depth, third-party mentions, and whether your brand has authority in the space. Schema reduces friction. It doesn't build trust from scratch.
A useful way to think about it: schema removes a reason for AI to skip you. It doesn't give AI a reason to pick you.
What Schema Actually Controls
Schema signals: content type, authorship, publication date, entity identity, Q&A structure. It makes your content easier to extract and verify. It's why the generative engine optimization work - building authority, earning third-party mentions, writing extractable content - still has to happen alongside the technical layer.
One quick test before you go implement anything: pull one paragraph from your page. If an AI surfaced just that paragraph as an answer, would it actually be useful? If not, the content needs work first. Schema can't rescue content that doesn't answer the question well.
How Do You Check If Your Structured Data Is Working for AI?
Get the implementation validated before you test citation impact. Otherwise you can't tell whether a lack of citations is a content problem or a broken schema problem.
Validation
Google's Rich Results Test catches syntax errors and tells you whether your schema qualifies for rich results. The Schema.org validator is more thorough - it checks field types, required properties, and flags incomplete implementations. Run both. The first catches formatting mistakes, the second catches the half-populated schemas that land you in the 41.6% citation rate range.
Seeing the Impact on AI Citations
After your schema is clean and validated, start querying ChatGPT, Perplexity, and Gemini with the questions your buyers would ask. Look for whether your content appears, where it appears relative to competitors, and whether the sentiment is positive.
Doing this manually for more than a handful of queries is tedious. SuperGEO handles this automatically - it scans your brand across the major AI engines on a schedule and tracks citation rate, mention sentiment, and competitor share of voice in one place. If you want a real picture of whether your structured data work is translating into visibility, that's a much faster way to track your AI visibility than querying engines by hand every few weeks.
FAQ
Does Structured Data Help with Perplexity and ChatGPT, Not Just Google?
Yes, though it works differently for each. Google's AI Mode directly processes schema during answer synthesis. Perplexity and ChatGPT with browsing rely on live web results - and those get indexed through Bing (in ChatGPT's case) or Perplexity's own crawler. Pages with clean schema tend to be indexed more completely, which means they're more likely to get pulled into responses.
The direct mechanism is clearest with Google. But schema's downstream effects - better indexation, cleaner entity signals, richer content extraction - improve your odds across all AI engines.
How Many Schema Types Should I Use on One Page?
Use as many as accurately describe the content. Don't add schema types that don't apply - that's the kind of thing that tips you into "generic/misleading" territory and hurts your citation rate.
In practice: a blog post typically warrants Article + FAQPage + BreadcrumbList. A product page: Product + Organization + AggregateRating. Three is a reasonable target to aim for, and the data supports it - pages with 3+ schema types have a 13% higher citation likelihood than those with one or two.
Do I Need a Developer to Implement This?
For most setups, no. JSON-LD sits in a <script> tag in your <head> - no HTML restructuring needed. WordPress, Webflow, and Framer all have plugins or built-in fields that handle this without touching code. If you're on a custom stack, it's a copy-paste with a few variable substitutions. The templates in this post are ready to use.
Schema is one of those things that looks simple and turns out to be fiddly in practice. The types are easy to understand. The failure mode - adding it half-finished and making things marginally worse - is less obvious until you see the citation data.
Start with the Tier 1 types, fill every field, and validate before you ship. Then give it a few weeks and actually check what's happening in AI search responses. The technical work is maybe a few hours. But it only pays off if you do it properly and then measure whether it's working.
If you want to skip the manual querying part, SuperGEO's free audit shows you exactly where you stand across ChatGPT, Perplexity, and Google AI Overviews before you spend any time optimizing.