Somebody adds structured data to a few hundred pages. The testing tool reports no errors. Search Console confirms the items are detected. Months later the search results look exactly as they did before, and the obvious conclusion is that the markup is broken somewhere.
Usually it is not broken. Validation and display answer different questions, and passing the first has never guaranteed the second.
Does valid structured data guarantee a rich result?
No. Validation checks that your markup is well formed and carries the properties a type requires. Whether a rich result appears is decided separately, at query time, by Google — and it has always been described as eligibility rather than entitlement.
This surprises people because every other technical check on a site is deterministic. A redirect either fires or it does not. A canonical tag either points somewhere or it does not. Structured data behaves more like a submission than a setting: you supply information in a machine-readable form and something else decides what to do with it.
What are the three gates?
Markup passes through three checks in sequence, and each one fails in a different place. Knowing which gate you are stuck at tells you whether there is any work to do at all.
| Gate | What it decides | Where you see the failure |
|---|---|---|
| Syntax and required properties | Is the markup readable and complete | Errors in the rich results test |
| Eligibility | Does this page and this type qualify at all | Item valid, no enhancement report appears |
| Display | Does this query deserve the richer format | Report says valid, the live result stays plain |
Only the first gate is fully in your hands. The second is partly — it depends on the type you chose and whether the page is genuinely the kind of thing that type describes. The third is not yours at all, and no amount of additional markup will change it.
If Search Console shows your items as valid with no errors and no warnings, the markup is done. Adding more properties to a page that already passes is effort spent on the one gate that was never the problem.
Does the markup have to match what the page shows?
Yes, and this is the most common real failure behind valid-but-invisible markup. Structured data is meant to describe content a visitor can see on the page. Markup that describes something absent from the page is a policy problem, not a syntax one, and a validator has no way to catch it.
The pattern shows up most often where markup is generated from a database rather than from the rendered page. A template emits review schema with an aggregate rating while the page itself shows no reviews. A product schema block carries a price the visible page no longer lists because it moved behind a quote form. Each validates perfectly. Each describes a page that does not exist.
The check is manual and takes a minute per template. Open the page, open its markup, and confirm that every value a user would care about — price, rating, availability, author, date — appears somewhere a visitor can read it.
Is the page the kind of thing the type describes?
Eligibility depends on the page being a genuine instance of the type, not merely carrying the fields. A category listing marked up as a single product, or a service page marked up as an article, satisfies a validator and fails the second gate quietly.
Two habits cause most of this. The first is picking a type because it produces an attractive result rather than because it describes the page. The second is applying one schema markup template sitewide, so every page claims to be whatever the template says. Both produce clean validation across hundreds of URLs and enhancements on none of them.
Adding a second and third schema type to a page that is already valid, on the theory that more signals must help. It gives a search engine more contradictory claims to reconcile about what the page is, and contradiction is a reason to show less, not more.
How long should you wait before concluding it failed?
Long enough for the pages to be recrawled and the enhancement reports to populate, which for most sites is a few weeks and for large or rarely crawled sites can be longer. A report that is empty three days after deployment has told you nothing.
Watch two things while you wait. Whether the items are being detected at all, which tells you the pages are being fetched and parsed, and whether the count of valid items tracks the number of pages you marked up. A detected count stuck far below the page count usually means a rendering problem — markup injected by client-side script that is not present in what gets fetched.
What is worth fixing, and what is not?
Fix anything at the first gate, because errors there can suppress the whole item. Fix any disagreement between markup and page, because that one carries risk beyond a missing rich result. Fix a type that does not describe the page, since it was never going to qualify.
Do not keep adding properties to items already reported as valid, and do not rebuild the markup because a competitor's result looks richer than yours. Their advantage is more likely to be the query, the site, or the result type being available in their category and not in yours.
Open Search Console, find whether your items are valid, and decide from that alone whether you have a markup problem or an expectation problem. They need very different amounts of your time.