Google's own documentation says structured data is not a ranking requirement. No schema field adds positions on its own. What it does change is how a machine reads your page: which entity you are, what the page describes, and whether the page can appear as a rich result. So schema markup changes interpretation and eligibility, not authority.
Structured data, schema markup and JSON-LD are three different things, and mixing them up causes most of the confusion.
Structured data is any format that adds explicit labels to content so a machine can parse it without guessing. Schema.org is the shared vocabulary of types and properties those labels use. JSON-LD is one way to write that vocabulary, as a script block in the page head.
People treat the three words as synonyms. They are not. You can publish structured data as Microdata or RDFa, and crawlers still read both. Google's documentation recommends JSON-LD, and it survives template changes most cleanly.
A plain definition: structured data is a machine-readable statement about a page's content, separate from the visible text, that a crawler can parse and treat as a claim about that page. The word claim matters. You assert something, and the search engine decides whether to use it.
| Term | What it is | Who defines it | Practical effect |
|---|---|---|---|
| Structured data | Any labelled data format a crawler can parse | Concept, no single owner | Makes content explicit rather than inferred |
| Schema.org vocabulary | Shared list of types (Article, Product) and properties (author, price) | schema.org, founded by Google, Microsoft, Yahoo and Yandex | Gives you a common language crawlers already understand |
| JSON-LD | JavaScript Object Notation for Linked Data, one serialisation | W3C JSON-LD 1.1 Recommendation | Format Google documents as its preferred option |
| Microdata / RDFa | Older inline markup formats | WHATWG HTML / W3C | Still parsed, harder to maintain |
| Rich result | A search listing with extra visual elements | Google Search Central | Eligibility only, never assured |
Schema markup buys rich result eligibility, not ranking positions.
Google's Search Essentials states that structured data is not required for inclusion and does not by itself improve ranking. The visible wins are eligibility for enhanced listings: review stars, breadcrumb paths, product availability, article bylines.
Ranking movement, when it happens, usually comes from a side effect. Clearer entity signals can help Google match a page to a query intent it previously misread. That is an interpretation gain, not a ranking factor. Treat any claim that schema "boosts rankings by X percent" as unsourced unless it names the test, the sample and the date.
There is a real eligibility change to plan around. Google's Search Central documentation records that HowTo rich results were deprecated and stopped showing in 2023, and FAQ rich results were restricted to well-known authoritative government and health sites. A FAQPage block on a B2B site is still valid markup, but in most cases it will not produce a dropdown.
Entity clarity is the quiet win: it tells the machine which thing you are.
An entity, in search terms, is a distinct thing a search engine can identify: a company, a person, a product, a place. Entity clarity means the machine can tell which thing you are and how your pages relate to it. Schema is one input, alongside consistent naming, an About page, and external references.
JSON-LD helps here because it lets you state relationships explicitly. An Organization node with a sameAs property pointing to a Wikidata or LinkedIn URL tells the crawler that this page and that profile describe the same entity. Without it, the crawler infers, and inference is where mismatches creep in.
Google's documentation on Organization structured data describes sameAs as a way to identify the organisation's presence on other sites. That is the concrete mechanism. It does not promise a knowledge panel, and Google publishes no figure for how much schema contributes to entity resolution.
Structured data helps AI systems read your page, but nobody has published how much it helps citations.
AI answer engines parse pages with the same kind of HTML parser a crawler uses, so they can read JSON-LD. Whether a given system uses it for retrieval, ranking or citation is not documented publicly by OpenAI, Anthropic or Google for their answer surfaces.
What is documented: OpenAI publishes a crawler called GPTBot and a separate OAI-SearchBot for search-related crawling, both controlled through robots.txt. That tells you about access, not how the parsed schema is weighted.
The honest position is that structured data removes ambiguity. If an AI system is trying to work out whether your page is a product page, an article or a support doc, explicit markup answers that question in one parse. Whether that changes citation behaviour is unmeasured in public.
JSON-LD beats Microdata for a B2B site, mostly because it breaks loudly instead of silently.
Google's structured data documentation names JSON-LD as its recommended format. It sits in one script block, it does not tangle with your HTML, and a developer can generate it from a template or CMS field.
Microdata and RDFa still work and Google still parses them. The reason to move is maintenance. Inline markup breaks when a theme or component changes, and broken markup silently stops qualifying for anything. A JSON-LD block in the head fails visibly and is easier to test.
A minimal Article example looks like this: a script tag with type application/ld+json, containing @context set to https://schema.org, @type set to Article, plus headline, datePublished and author properties. That is enough to be valid. Add more only when you have real values to fill.
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Your page title",
"datePublished": "2026-09-29",
"author": { "@type": "Person", "name": "Author Name" }
}
</script>For a B2B site, schema pays off on product, organisation, breadcrumb and article pages, and pays off almost nowhere else.
It matters when the markup maps to a result type that still displays. Product pages with price and availability, organisation identity, breadcrumbs, article authorship and video content all still have live rich result documentation in Google Search Central. If your page type is on that list, the work has a visible payoff.
It matters less when the page is a long-form explainer with no matching result type. Marking it up as Article is good hygiene and helps machines parse the author and date. It will not change the listing. That is fine, as long as nobody sold it to you as a ranking lever.
It matters almost not at all if the underlying content is thin. Schema describes content; it does not create it. A Product node on a page with no specifications, no price and no availability gives the crawler nothing true to say.
Three checks tell you whether your structured data actually works, and they run in a fixed order.
First, validate the syntax with the Schema Markup Validator at validator.schema.org. It checks against the vocabulary, not against Google's rich result rules. Second, test eligibility with Google's Rich Results Test, which reports which result types the page qualifies for.
Third, watch Search Console. The Enhancements reports list valid items, warnings and errors by type, and they update as Google recrawls. A page can pass validation and still show zero rich results, because eligibility is not a promise.
A self-check list you can run this week:
- Does every page type you care about have a JSON-LD block, or none at all?
- Do the values match the visible page text, including price and date?
- Is the organisation name identical across schema, footer and About page?
- Does sameAs point to profiles that actually exist and are controlled by you?
- Are deprecated types still sitting in the template, doing nothing?
- Does Search Console show errors on a page type you thought was clean?
Key Facts at a Glance
- Google's Search Essentials states structured data is not a ranking requirement.
- JSON-LD is Google's documented recommended format; Microdata and RDFa are still parsed.
- HowTo rich results were deprecated in 2023; FAQ rich results are restricted to authoritative government and health sites.
- Schema Markup Validator checks the vocabulary; Rich Results Test checks Google eligibility.
- sameAs links a page's entity to external profiles; it does not promise a knowledge panel.
Does schema markup actually improve rankings?
Not by itself. Google's Search Essentials says structured data is not a ranking requirement. The measurable change is eligibility for rich results. Any ranking lift usually comes from clearer interpretation of what the page is about.
What is the difference between schema and JSON-LD?
Schema.org is the vocabulary of types and properties. JSON-LD is one file format for writing that vocabulary into a page. You can express the same schema in Microdata or RDFa instead, but Google documents JSON-LD as its recommendation.
Which schema types still show rich results?
The ones with live documentation in Google Search Central: Product, Article, Breadcrumb, Organization, Video, Event, Recipe and a few others. FAQ and HowTo have been restricted or deprecated, so check the current page before you build around them.
How do I test if my structured data works?
Validate syntax at validator.schema.org, then check eligibility with Google's Rich Results Test, then watch the Enhancements reports in Search Console. All three, in that order. Validation alone does not mean a result will display.
Does structured data help with AI search?
It removes ambiguity, because AI systems parse HTML and can read JSON-LD. Whether a specific answer engine weights it for citation is not publicly documented. Treat the access question (robots.txt, crawler tokens) as the one you can actually control.