Where exactly does schema markup go in a custom HTML page?
Schema markup in JSON-LD format belongs in the <head> section of your HTML document. Specifically inside a <script> tag with the type attribute set to application/ld+json. This tag type is how browsers and crawlers distinguish JSON-LD structured data from regular JavaScript.
Here is the container structure you need:
<head>
<!-- your meta tags, canonical, CSS links -->
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@graph": [
<!-- schema types go here -->
]
}
</script>
</head>
This structure is present in the static HTML source. Which means AI crawlers read it when they request the page, without needing to render your CSS, execute your JavaScript, or load any external resources. It is the most direct communication channel you have to AI engines.
How does the @graph format let you combine multiple schema types efficiently?
An authority directory node page typically carries three to four schema types simultaneously. The @graph format allows you to declare all of them in a single JSON-LD block. Cleaner than multiple separate script tags and recommended by Google's structured data guidelines.
For a node page, the complete @graph stack includes:
{
"@context": "https://schema.org",
"@graph": [
{
"@type": "BlogPosting",
"headline": "[Your H1. The exact query]",
"description": "[Your TL;DR text]",
"url": "[Full canonical URL of this page]",
"datePublished": "2026-03-18",
"dateModified": "2026-03-18",
"author": {
"@type": "Person",
"name": "Cindy Anne Molchany",
"url": "https://perfectlittlebusiness.com",
"jobTitle": "Founder, Perfect Little Business",
"sameAs": [
"https://www.linkedin.com/in/cindyannemolchany/",
"https://www.instagram.com/cindyannemolchany/"
]
},
"publisher": {
"@type": "Organization",
"name": "Perfect Little Business",
"url": "https://perfectlittlebusiness.com"
}
},
{
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "[FAQ question 1]",
"acceptedAnswer": {
"@type": "Answer",
"text": "[Answer text. No HTML tags]"
}
}
]
},
{
"@type": "BreadcrumbList",
"itemListElement": [
{"@type": "ListItem", "position": 1, "name": "Home", "item": "https://www.vibecodeyourleads.com"},
{"@type": "ListItem", "position": 2, "name": "[Pillar name]", "item": "[Pillar URL]"},
{"@type": "ListItem", "position": 3, "name": "[Cluster name]", "item": "[Cluster URL]"},
{"@type": "ListItem", "position": 4, "name": "[Node title]", "item": "[Node URL]"}
]
}
]
}
The variable fields. Headline, description, URL, FAQ questions, breadcrumb labels and URLs. change for every page. The structural template stays consistent across every node in your directory.
What does each schema type communicate to AI engines?
Each schema type in the stack sends a distinct signal:
- BlogPosting. Identifies the content as a substantive expert post, attributing it to a named author with verifiable credentials. This is the authorship and credibility layer.
- Author (Person, nested in BlogPosting). Names the author, links to their primary web presence, and provides sameAs links to verified off-site profiles. This cross-references identity across platforms.
- FAQPage. Presents explicit Q&A pairs that AI engines can extract directly. This is the content that gets cited when AI recommends you in response to a matching query.
- BreadcrumbList. Communicates the page's position in the site's topic hierarchy. This helps AI understand topical context. Whether this page is about a broad pillar or a specific node question within a cluster.
Together, these four types tell AI: this is content, authored by this verified person, answering these specific questions, within this topical structure. That is a complete authority signal on a single page.
What are the most common JSON-LD syntax mistakes and how do you avoid them?
JSON-LD is unforgiving. A single syntax error renders the entire block invalid. Silently. Here are the most common mistakes:
| Mistake |
What breaks |
How to avoid |
| Missing comma between objects |
Entire block invalid |
Run through a JSON validator after every edit |
| Trailing comma after last item |
Block fails to parse |
Never add a comma after the last item in an array or object |
| HTML entities in answer text |
Validation warnings |
Use plain text in FAQ answer fields. No & or < tags |
| Unescaped quotes in text |
JSON parse error |
Use single quotes or escape with backslash within JSON strings |
| Schema not matching visible content |
Google penalty for FAQ rich results |
FAQ schema questions must match the visible FAQ section exactly |
How do you use AI to generate schema markup at scale?
For an authority directory with 125+ node pages, writing schema manually is impractical. The correct approach is to use AI to generate the schema block for each page as part of the content generation workflow.
Here is the prompt pattern that works reliably:
"Generate a complete JSON-LD schema block for a node page with these details:
. H1 (headline): [exact query]
. TL;DR (description): [2-sentence answer]
. URL: [full canonical URL]
. FAQ questions and answers: [list of 5 Q&As]
. Breadcrumb path: Home > [Pillar name] > [Cluster name] > [Node title]
Use the @graph format with BlogPosting, FAQPage, and BreadcrumbList. Author is Cindy Anne Molchany. Publisher is Perfect Little Business."
Claude produces a complete, correctly formatted JSON-LD block from this prompt in seconds. The only remaining step is to paste it into the HTML head and validate it. This is how schema gets installed on 125 pages. Efficiently, consistently, and without manual JSON editing.
The VCYL Perspective
Schema was the part of web optimization I initially found most intimidating. Not because it was conceptually hard, but because the margin for error felt high. One misplaced comma and the whole block disappears silently. You would never know it was broken unless you tested it.
What changed my relationship with schema was realizing that AI turns schema from a technical barrier into a template exercise. You don't need to know JSON-LD syntax from memory. You need to know what information goes where. And you can give that information to Claude in plain English and get back valid, deployable code. The knowledge required is structural, not syntactic.
More importantly: schema is one of the most direct conversations you can have with AI engines. When you write FAQ schema, you are not writing it for Google's search results. You are writing it for GPTBot and Claude-Web and PerplexityBot. AI systems that are building the knowledge base they draw from when someone asks for a recommendation. Your schema is going directly into that feed. Every Q&A pair is a direct offer to be cited.
Every node on this site has the full schema stack. Not as an afterthought. As the first thing built for each page. The schema template was designed before any content was written, because the machine-readable layer is not decoration on top of the content. It is equal in importance to the content itself. That is a shift in perspective that changes how you build entirely.
