The most important thing: Model the real business and visible content accurately; keep the markup maintained.
Related play: Publish a business context page — a factual reference with service fit, comparisons and a practical 2027 v1 implementation.
1Choose the right structured-data type
Use Organization or a more specific applicable subtype. A local business, online store, product and article are different things. Model actual branches separately when needed, with stable identifiers and clear relationships. Do not label everything LocalBusiness just to pursue local visibility.
The selected types describe what exists in the real world.
2Use stable identifiers and genuine profiles
Give the organization a consistent @id, such as https://example.com/#organization. Provide applicable name, URL, logo and contact information. Use sameAs for official profiles of the same business or person; do not use it for competitors, articles about a topic or brands you merely sell.
Identifiers and external profile references are consistent and meaningful.
3Match the visible page
Mark up actual products, offers, articles or business information only when appropriate. Keep prices, stock, hours and contact information synchronized with the page and operational systems. Never invent ratings, reviews or certifications for markup.
A reviewer can find the same facts in the visible page or connected business record.
4Validate and monitor
Parse the JSON, use an applicable structured-data validator and test supported Google features with Rich Results Test. Resolve errors; evaluate recommendations in context. Keep validation in the release checklist and check live pages after changes.
The live markup is valid and maintained, with no unsupported public claims.
A concrete example
Avoid this mistake: There is no special universal “AI schema” that guarantees recommendations. A valid schema.org object is not the same thing as eligibility for a particular Google rich result.