ComparisonUpdated 2026-09-21

Structured data versus visible text on a product page

Structured data lets a machine read identifiers without parsing prose; visible text is what a model quotes and what survives when markup is ignored. You need both, and they must agree — when markup says one capacity and the text another, the page contradicts itself.

Side by side

| | Structured data (JSON-LD) | Visible text | | --- | --- | --- | | Best at | identifiers, prices, precise fields | facts a buyer asks about, in sentences | | Read by | parsers, search engines, some AI pipelines | every reader, and every model that quotes | | Survives | only if the consumer parses it | whatever is fetched | | Weak at | nuance, limits, "does not include" | exact machine-readable fields |

When they disagree

A common failure: the template's markup is generated from last season's catalogue while the text was updated by hand. The page now states two capacities. A model has no way to know which one the seller meant, and a reviewer checking the page sees the right one — so the error is found only in answers.

The rule

Generate both from the same source, in the same release, and check that the identifiers in the markup appear verbatim in the text. The page-level checklist is at /checklists/product-page-ai-readiness; what a published fact can and cannot prove is at /evidence.

Sources

Related

← All pages