Model the visible facts first
Identify whether the page primarily represents a service, product, article, organization, or collection. Mark up facts supported by the content rather than selecting the most impressive-looking type. A company’s software product is not the company itself, and a publisher’s local directory is not an identity alias for that publisher. Keep those entities distinct.
- ✓ Choose types that fit the page and its evidence.
- ✓ Do not add fabricated reviews, ratings, prices, or credentials.
Reuse identifiers where the entity recurs
An absolute identifier can connect the same organization or product across pages. Use references consistently and check that linked nodes exist in the intended graph. Separate the page identifier from the thing described by the page. SchemaGraph can help inspect repeated identities and unresolved references; suggestions still need validation against the site’s actual entity model.
- ✓ Give recurring entities stable, controlled identifiers.
- ✓ Check duplicate nodes and conflicting names across templates.
Validate meaning as well as syntax
A syntactically valid block can still describe the wrong entity or contradict a page. Review visible content, supported properties, and any search-feature eligibility requirements separately. Structured data does not guarantee rankings, rich results, or AI citations. Revalidate after template changes and preserve the previous version so a regression can be traced and corrected.
- ✓ Test rendered output as well as the source template.
- ✓ Maintain an inventory of schema owners and affected page types.