JSON-LD is a script block that describes a page's content using schema.org types. For a data vendor, the useful core is Organization for the company, Article for resources, FAQPage for visible FAQs, BreadcrumbList for navigation and Dataset for data products. It makes pages eligible for rich results and easier for AI systems to parse, but it guarantees no ranking.
How to use this article
Read the checklist here, then use the linked hub and product pages for procurement citations.
A data vendor's website is read by two audiences that rarely meet: buyers comparing coverage and delivery terms, and machines deciding what each page is about. JSON-LD is the part of the page written for the second audience. Done well, it is a short, accurate description of what is already visible. Done carelessly, it becomes a second, quieter version of the page that drifts away from the first.
Why structured data matters for a data vendor
Data products are hard to describe in prose alone. A product page might mention a refresh cadence, a delivery format, a geography and a licensing model, scattered across paragraphs and tables. A crawler can extract those facts, but it has to guess at their relationships. Structured data states them directly: this page is a Dataset, this is its name, this is its description, this organization publishes it.
That matters more now that AI assistants and retrieval systems summarize vendor pages for buyers. A model that reads a clean Article block knows the headline, the publisher and the publication date without inferring them from layout. That does not guarantee a citation, and nobody can promise one, but it removes ambiguity from the facts you most want repeated correctly. The broader readiness picture, covering crawl access and server-rendered HTML, is in our AI search readiness guide.
The second benefit is eligibility for rich results in Google Search. Google's structured data introduction is explicit that markup makes a page eligible for enhanced display and does not guarantee it will appear. Treat structured data as accurate labeling, not a ranking lever. If someone sells it to you as a ranking boost, ask them for the source.
The core schema.org types a data-vendor site needs
Most data-vendor sites need fewer types than the schema.org vocabulary suggests. Start with the set below and add others only when a page genuinely fits them.
Page types on a data-vendor site and the schema.org types that describe them. Source: schema.org type definitions, as of November 2026.
Page type
schema.org type
What it states
Sitewide (usually the home page)
Organization, optionally LocalBusiness
Legal name, logo, URL, contact points and, for LocalBusiness, a real street address
Blog post or resource
Article
Headline, author, publisher, image, publish and modified dates
Any page with a visible FAQ
FAQPage
The questions and answers shown on the page, verbatim
Every page below the home page
BreadcrumbList
The navigation path from home to the current page
Data product page
Dataset
Name, description, creator, coverage and distribution details
Service page (enrichment, matching, onboarding)
Service
Service type, provider and area served
Organization. Emit it once, with a stable @id such as your domain plus #organization, and reference that @id as the publisher everywhere else. That keeps one canonical description of the company instead of dozens of slightly different copies. Add LocalBusiness only if you actually serve visitors at a physical address you are willing to publish.
Article. Every resource post should carry it. Google's Article documentation lists the properties it reads, including headline, image, author and dates. The author should be the real accountable party, which for many vendors is the organization rather than a named individual.
FAQPage. Use it only where the questions and answers are rendered on the page. Google has narrowed which sites receive FAQ rich results in search, so value FAQPage first as clean machine-readable structure rather than a rich-result guarantee. Our FAQ schema patterns guide covers how to write questions buyers actually ask.
BreadcrumbList. Cheap to generate from your routing and useful on almost every page. It tells crawlers that a resource sits under Resources, or that a product sits under Products, which helps a data catalog read as a catalog.
Dataset. This is the type most specific to a data vendor and the most often skipped. A product page describing a licensable file or feed is a reasonable Dataset candidate. A mobile advertising ID (MAID) device feed, such as GSDSI's MAID feed, is a typical example. At minimum supply a name and description; add temporal coverage, spatial coverage and distribution format only where the page states them. The Dataset block summarizes the product, while a published data dictionary documents each field and its meaning, so keep the two in agreement.
Service. Use it for work you perform rather than files you license, such as identity matching or list hygiene. Do not mark a data file as a Service to avoid filling in Dataset fields.
As a working example, this site runs Article, BreadcrumbList and FAQPage together on resource posts, emits FAQPage only where a visible FAQ block exists, uses Dataset on data-product pages, and emits Organization once sitewide. A build-time script checks that each Article and Dataset block carries the fields Google requires, so a template change cannot silently drop them. You can browse those product pages in the data catalog.
Common structured-data mistakes
Marking up content that is not on the page. An FAQPage whose answers appear nowhere in the HTML, or a Dataset description longer and more generous than the visible copy, is describing a different page.
Duplicate and conflicting blocks. A CMS plugin emits one Article, a theme emits another, and the two disagree on headline or author. Pick one source of truth and remove the other.
Stale or inconsistent dates. If the visible byline says one date and dateModified says another, a reader and a crawler reach different conclusions about freshness. Generate both from the same field.
Claiming credentials nobody holds. Adding reviewedBy or a named expert author because it looks authoritative, when no such review happened, is a false statement in machine-readable form. Leave the property out until it is true.
Using a type that does not match the content. A blog post marked as a Product, or a sales page marked as a Dataset with no data described, tells parsers something the page does not support.
How to validate JSON-LD
Validation has three layers, and each catches something the others miss.
Google's Rich Results Test shows which rich-result features a URL is eligible for and flags missing required properties for the types Google supports.
The Schema Markup Validator checks your markup against the full schema.org vocabulary, including types such as Dataset and Service that may not produce any rich result at all.
Read the rendered HTML yourself. View the page source of the deployed page, find each JSON-LD script and compare it with what a visitor sees: headline, dates, author, FAQ wording, breadcrumb path. Neither tool can tell you whether the markup is honest.
The third step is the one teams skip and the one that matters most. Both validators will pass a perfectly formed block that describes the wrong page. If your site prerenders HTML, check the prerendered output rather than the client-rendered DOM, since that is what many crawlers fetch first. Once the manual check passes, automate the required-field portion in your build so a regression fails before it ships.
Rolling it out without making a mess
Start with your highest-traffic page type, which for most data vendors is the resource library or the product catalog. Get Article or Dataset right there, validate a handful of pages by hand, then extend the same template to the rest. Add BreadcrumbList next, since it is mechanical, and FAQPage only where an FAQ already exists.
Keep one script per concern and generate every value from the same data that renders the visible page. When the content changes, the markup should change in the same commit, without a separate step someone has to remember. That is the whole discipline: structured data is a faithful summary of the page, kept in sync, and nothing more.
Google does not promise a ranking benefit from structured data. Its documentation says markup can make a page eligible for rich results, with no guarantee they will appear. The practical value is accurate, machine-readable description of the page, which also helps AI systems parse it correctly.
What is the minimum schema a data vendor site needs?
Organization once sitewide, Article on every resource post and BreadcrumbList on pages below the home page. Add FAQPage where a visible FAQ exists and Dataset on pages that describe a licensable data product. Everything else is optional until a page genuinely fits it.
How do you validate structured data?
Run the page through Google's Rich Results Test for supported features, the Schema Markup Validator for the full schema.org vocabulary, and then read the rendered HTML to confirm every JSON-LD value matches what a visitor sees. The validators check form; only the manual read checks honesty.
What is the most common structured-data mistake?
Marking up content that is not visible on the page, such as FAQ answers that never render or descriptions that differ from the visible copy. Close behind are duplicate blocks from competing plugins and dates in JSON-LD that disagree with the visible byline.
Should a data product page use Dataset or Product?
Use Dataset when the page describes a file or feed: its contents, coverage and distribution. Product is built around offers, reviews and ratings, and Google's Product structured data documentation lists those as required or recommended properties. If the page has no published price or reviews, Dataset is usually the more honest fit.
✓ Opt-Out Request Honored via Global Privacy Control
We use cookies and similar technologies to improve your experience. No marketing cookies are pre-selected. Learn more in our Privacy Policy. Your Privacy Choices.