Nothing about the markup changed; the things around it did. Structured data depends on the template that prints it, the data that fills it, and Google's current eligibility rules, and all three move without anyone editing a line of schema.
Why does schema stop working when nobody touched it?
Because almost no one writes schema by hand. It is output: a template reads fields from a database or a plugin and prints the markup on every page of that type. Change the input and the output changes with it.
That makes schema a dependency rather than a document. A product goes out of stock and the availability value flips. An editor clears a field that was feeding a required property. A plugin update renames a key. None of those look like schema work to the person who did them, and none of them show up in a release note that mentions markup.
What breaks first?
Field-level data, almost every time. The template keeps printing valid syntax, so the markup still parses; it just describes something that is no longer true or is now empty.
| What changed | How it shows up | Where you see it |
|---|---|---|
| A field was emptied or renamed | A required property is missing; the item becomes ineligible | Search Console enhancement report, as invalid items |
| Price, stock or rating data moved | Markup contradicts the visible page | Manual spot check, or a sudden drop in eligible items |
| A template was rebuilt | The markup block was not carried over at all | Enhancement report loses the page type entirely |
| A plugin or CMS update | Duplicate or conflicting blocks on one page | Rich Results Test shows two of the same type |
| Google changed eligibility | Markup is still valid; the rich result is gone | Valid items hold steady, appearance drops to zero |
The last row is the one teams misdiagnose most, because every tool says the implementation is fine.
Valid and eligible are two different states, and only one of them is yours to control. A validator tells you the syntax is correct. Whether a correct item is still shown is Google's decision, and it has changed that decision several times. Track appearance, not validity, or you will spend a sprint fixing markup that was never broken.
Has Google really taken rich results away from valid markup?
Yes, more than once, and it announced each one. In 2023 Google narrowed FAQ rich results to authoritative government and health sites and cut HowTo results from desktop and mobile. The markup on every other site stayed valid and stopped producing anything.
Review snippets have the same history. Google restricted self-serving reviews, the kind a business marks up about itself on its own organization or local business page, and sites that had built review schema around that pattern lost the stars while keeping a clean validation report.
The lesson is about where to put your effort, not about abandoning schema. Types tied to a specific visual treatment can lose that treatment on Google's schedule. Types that describe what a page is, so a machine can read it without guessing, keep earning their place regardless. There is a longer argument for that in why schema markup is worth shipping without a rich result.
Does a template rebuild always lose the markup?
Often enough that it belongs on every migration checklist. Schema markup is invisible on the rendered page, so a redesign can be reviewed by five people and shipped without anyone noticing the block is gone.
The pattern is predictable. A new template is built from a fresh component, the old one carried a hand-added script block that was never documented, and the markup leaves with the file. Nobody sees it because the page looks right. The first signal arrives weeks later, when the enhancement report shows zero valid items for a type that used to have thousands.
Product schema on ecommerce templates is the most expensive version of this, because the markup is doing real work in the results and there are usually several templates involved. Check the schema output of every page type you rebuilt, not just the one you were testing on.
Testing one URL and calling the type healthy. The Rich Results Test checks the page you give it. A field that is empty on four percent of your products produces a clean test on the URL you picked and thousands of invalid items across the catalog. Read the count in the enhancement report before you read any single-URL test.
How do you find a break before it costs you anything?
Watch counts over time rather than running audits. Every failure mode above changes a number in the Search Console enhancement report for that type, and the direction tells you which failure you have.
Three patterns cover most of it. Valid items falling while invalid items rise means a data or required-property problem. Both falling together means the markup stopped being printed, which points at a template or plugin change. Valid items holding flat while impressions of the rich result fall means eligibility changed, and no amount of markup work will bring it back.
One check belongs in the release process rather than the monthly review: after any change to a template, a theme, or a plugin, pull the rendered schema from one live URL of each affected page type and compare it against the previous version. That is a five-minute diff that catches the silent removal case, which is the one that takes longest to notice on its own. The measurement side of this is covered in how to measure whether your structured data is actually working.
What do you do about markup that is valid but doing nothing?
Decide whether it still has a job. Schema that no longer produces a rich result may still describe the page usefully for anything parsing it, including the systems generating AI answers, and it costs nothing to leave in place while it is accurate.
Inaccurate markup is the opposite. Schema that contradicts the visible page is a manual action risk and is worth removing the same week you find it, not queuing. When to remove structured data covers where that line sits.
Which page type do you check first?
The one whose markup appears in the results and whose data changes daily. For most sites that is products or local pages, and the honest question is not whether your schema validates but when anyone last looked at the count. If the answer is launch day, your organic search listings are running on markup nobody has verified in months, and the only way to know what it says is to read what the template prints today.