Schema markup that actually earns citations
Five graphs carry almost all the value for a knowledge site: Organization, WebSite, BreadcrumbList, Article or FAQPage, and Person for the author. Connecting them with stable identifiers matters more than adding a sixth type.
Most structured data advice is a catalogue of available types. What matters more is that the graphs you do publish are connected, consistent and true, because a machine reading them is trying to build a picture of an organisation rather than to tick off types.
The five that carry the weight
| Graph | Says | Where |
|---|---|---|
| Organization | Who publishes this, where they are, how to reach them | Declared once, referenced from every page |
| WebSite | This domain is a thing with a name and a publisher | Home page |
| BreadcrumbList | Where this page sits in the structure | Every page below the top level |
| FAQPage or Article | What this specific page is and what it settles | The page itself |
| Person | Who wrote it and why they would know | On the page, referenced from the Article |
Connection is the part that gets skipped
Give each graph a stable identifier and reference it rather than repeating it. An Organization declared with an identifier at your domain root, referenced as the publisher of every article and the employer of every author, describes a coherent entity. The same information copied inline into forty pages describes forty loosely related documents.
Add the identifiers that connect you outward too, where they exist: your registry number, your entry in the relevant trade register, your profiles. This is how a machine works out that the company in the markup is the company it has read about elsewhere.
Structured data is not a checklist of types. It is a description of one organisation.
What to leave alone
- Review and rating markup on your own pages about yourself. It invites scrutiny, it is frequently disregarded, and the downside is a manual action rather than a lost rich result.
- HowTo markup on things that are not procedures. Rich result support has been withdrawn in most contexts and the type is often applied to marketing copy.
- Speakable, unless you genuinely have a voice surface and have read the current restrictions.
- Any type whose fields you have to invent values for. An invented field is a wrong field.
Author markup deserves more care than it gets
An author is a person with a name, a role, a page describing their experience, and ideally identifiers elsewhere. "Marketing Team" as an author is a missed opportunity on a knowledge page, because the whole reason first-hand sources are preferred is that somebody with specific standing said the thing.
Questions this settles
No. Structured data helps a machine interpret a page; it is not a ranking input you can accumulate. Adding types a page does not warrant makes interpretation worse.
JSON-LD, in the page source. It is the format search providers recommend, it is separable from the markup, and it is the one that survives a front-end redesign.
In the HTML. Script-injected markup is missing exactly when a page is fetched without a browser, which is the case for much machine reading.