Implementing Schema.org: Match DefinedTerm, FAQPage and HowTo to Visible Content
Choose a semantic type from the content you already publish, keep JSON-LD consistent with the visible page, and separate vocabulary from search-feature eligibility. The examples are implementation exercises, with no citation guarantee.
By
Tenten AI FDM Team
Forward-Deployed Marketing
Published
Updated
Read time
4 min

Structured data can describe a term, a set of questions or a procedure. Choosing among those types starts with what the page actually contains. It does not start with a promise that adding a particular type will make an AI system cite the page.
Google’s AI optimization guide says special Schema.org markup is not required for its AI search experiences. That guidance applies to Google; it does not establish how every other engine consumes structured data. Use markup to represent existing content accurately, alongside ordinary crawlability, useful text and reliable sources.
Choose a type before writing the JSON
| Visible content | Candidate vocabulary | Check before using it |
|---|---|---|
| A named term with a definition | DefinedTerm | The definition is actually present and appropriately scoped. |
| Publisher-written frequently asked questions and answers | FAQPage | Each marked question and answer is visible to readers. |
| Instructions for completing a task in steps | HowTo | The page provides the actual procedure, rather than a promotional list. |
These are semantic choices, not a ranking of citation performance. Schema.org defines a vocabulary. Search products decide separately which features they support and under what conditions.
DefinedTerm: publish the definition you mark
DefinedTerm describes a term and its definition. For this example, the visible glossary entry is:
Request-local cache: A cache whose stored values are reused only within one request.
The corresponding minimal JSON-LD is:
{
"@context": "https://schema.org",
"@type": "DefinedTerm",
"name": "Request-local cache",
"description": "A cache whose stored values are reused only within one request."
}
Use a real term set or stable URL only when it exists. Do not invent a glossary endpoint, authority relationship or organizational credential to complete a template. A definition also does not establish that the term is preferred by an AI system.
FAQPage: preserve the actual question and answer
The visible example is:
Does valid JSON-LD guarantee an AI citation? No. Valid JSON-LD does not guarantee an AI citation.
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Does valid JSON-LD guarantee an AI citation?",
"acceptedAnswer": {
"@type": "Answer",
"text": "No. Valid JSON-LD does not guarantee an AI citation."
}
}
]
}
FAQPage remains part of the vocabulary. Google’s update record says FAQ rich results stopped appearing on May 7, 2026; the related documentation was removed on June 15. Do not sell this example as eligibility for that retired Google feature. Do not add unnecessary FAQs merely to insert markup.
HowTo: mark a real sequence
HowTo describes instructions. This example’s visible procedure is Inspect existing JSON-LD:
- Open the rendered page: Inspect the HTML returned for the intended URL.
- Compare the content: Check that the marked statements match the visible page.
{
"@context": "https://schema.org",
"@type": "HowTo",
"name": "Inspect existing JSON-LD",
"step": [
{
"@type": "HowToStep",
"name": "Open the rendered page",
"text": "Inspect the HTML returned for the intended URL."
},
{
"@type": "HowToStep",
"name": "Compare the content",
"text": "Check that the marked statements match the visible page."
}
]
}
Do not turn a service’s marketing stages into instructions a reader can supposedly execute. Details such as time, supplies or costs belong in markup only when the real procedure supports them.
Keep implementation proportional
These fenced examples are teaching material, not additional live page entities. In an application, serialize the data with a JSON serializer and handle the HTML embedding boundary safely. Avoid concatenating untrusted strings into a script element. Check one representative output for each affected template, then rerun only after a relevant change or failure.
Google’s structured-data policies require accurate representation of visible content. Syntax validation, vocabulary validation and current search-feature checks answer different questions. None proves citation, traffic or conversion lift. If the existing content does not fit a useful type, keep the content clear and leave the extra markup out.
Summarize this page with AI
Related posts

Start with one workflow.
Agree on the next step.
Leave your contact details and question type. We will follow up by email to understand the work and discuss scope.