Structured data was born for Google rich snippets. Today it has a more interesting second use: it lets a machine understand your business without ambiguity.
Why an AI assistant cares
When a model retrieves your page it does not see your design: it sees text. JSON-LD hands it pre-interpreted facts — name, category, price, coverage, author — with nothing to infer. That lowers the risk of being filed under the wrong category or confused with a similarly named company.
The types that actually matter
- Organization: legal name, trading name, logo, sameAs with your official profiles. This is the base of your entity.
- WebSite with SearchAction if you have internal search.
- Product / SoftwareApplication with offers: price, currency and availability. Assistants happily cite verifiable pricing.
- FAQPage: each question with a short answer. It is the format that best matches how a generative answer is assembled.
- HowTo for step-by-step processes.
- BreadcrumbList so the site hierarchy is legible.
Entity consistency: the detail almost nobody handles
The sameAs field is your identity statement. It should point to the same profiles, with the same name and description, across your site, LinkedIn, business listings and industry directories. If you are "Acme Software" in one place and "Acme Digital Solutions Ltd" in another, you are splitting your authority across two entities.
Common mistakes
- Marking up content that does not exist on the visible page.
- Stale prices in offers: worse than none at all.
- Duplicating Organization blocks across pages with conflicting data.
- Filling FAQPage with questions nobody asks instead of the ones in your forms and support chat.
How to prioritise
Start with Organization + Product on the home and pricing pages, then FAQPage on your highest-intent pages. Measure the mention rate before and after: if AI starts quoting your price or integrations verbatim, the markup is working.