Validation confirms your markup parses. It doesn't confirm anything uses it. Check for conflicting duplicate blocks from the theme and plugin both emitting Organization, orphaned @id references pointing nowhere, and schema claims that contradict the visible page. Any of those makes valid markup functionally invisible.
"It validates" is the least useful thing you can know
Run a page through a structured data validator, and you'll get a green result. Almost everyone does. Validators check syntax — is this well-formed JSON, are the required properties present, is the type real. They do not check whether the markup describes your page accurately, connects to anything, or gets used.
Three failures pass validation cleanly.
Failure one: duplicate nodes fighting each other
The most common by a distance. Your theme emits an Organization node. Your SEO plugin emits another. A page-level block adds a third. All three validate. All three describe the same firm slightly differently — a different name format, a different logo URL, one with a phone number and two without.
An engine trying to establish who you are now has three partially-conflicting assertions from the same page. The usual outcome isn't an error; it's dilution. Nothing is wrong enough to reject, and nothing is confident enough to rely on.
How to check: View source, search for application/ld+json. Count the blocks. Then search within them for "@type": "Organization". More than one is your finding. Do the same for WebSite and FAQPage — two FAQPage nodes on one URL is a genuine conflict.
Failure two: orphaned identifiers
Well-built schema is a graph, not a list. Nodes reference each other by @id: a WebPage says it isPartOf a WebSite, a Service names its provider, an Article names its author.
Those references only work if the target node actually exists somewhere. When it doesn't — because a URL changed, or a node was written but never emitted, or someone typed #organisation in one place and #organization in another — the reference resolves to nothing. The markup still validates. The graph is just disconnected.
How to check: Copy every @id value that appears as a reference (inside isPartOf, about, author, provider, mainEntity). Then check each one appears as a definition — a node whose own @id is that string. Any reference without a matching definition is an orphan. Watch for casing, trailing slashes, and HTTP versus HTTPS; those count as different strings.
We've found this on our own site, in the version of the file we thought was live.
Failure three: schema that contradicts the page
Structured data is supposed to describe what a reader sees. When it describes something else, the mismatch is the problem — not the syntax.
The version we see most: a price in schema that no longer matches the price on the page. A firm updates its fee on the visible page and forgets the JSON-LD, which then asserts a different number in the one format built to be read and quoted verbatim. We caught exactly this on a client site where the markup named the same tier twice at two different prices.
Also common: an FAQ node containing questions that don't appear on the page, and an author or date that contradicts the byline.
How to check: Read the JSON-LD as prose, alongside the page. Every factual claim in the markup should be findable in the visible content.
The thing that changed in May 2026
If you've been told FAQ schema will expand your search listing, that advice is now out of date.
On 7 May 2026 Google deprecated FAQ rich results — the expandable question-and-answer dropdown under a listing. They no longer appear for any site, including the government and health sites that retained eligibility after the 2023 restriction. Search Console reporting and the Rich Results Test FAQ check were dropped in June 2026; Search Console API support ends in August 2026. HowTo rich results went the same way earlier, removed from mobile in August 2023 and desktop in September 2023.
Two things follow, and the industry has mostly overshot in both directions.
FAQPage and HowTo are still valid Schema.org types. Google has said unused structured data causes no problems, and its documentation explicitly notes the markup can stay. There is no urgency to strip it out.
But they are no longer a Google SERP lever. If someone recommends FAQ schema to you on the basis that it will make your listing bigger, they're describing a feature that no longer exists.
Whether FAQ markup helps with AI retrieval is a separate and genuinely unsettled question. Google's own AI-features guidance says no special markup is required for AI Overviews or AI Mode, while also saying structured data should match visible content. Anyone stating confidently that FAQ schema drives AI citations is ahead of the evidence.
One practical consequence: workflows that relied on the Rich Results Test to check FAQ markup need a different tool, and any reporting pulling FAQ data from the Search Console API will start returning nothing this month.
What you can and can't conclude
Verified — Duplicate nodes, orphaned @id references, and contradictions between markup and page. All three are directly observable in your own source, and you can show the evidence.
Unconfirmed — Whether any engine is actually consuming your schema and acting on it. There is no report for this. Google's Rich Results Test tells you eligibility for specific visual features, not whether your entity data influenced anything. For AI engines, there is no equivalent tool at all. Claims that particular markup "gets you cited" are inference, and should be labeled as such.
Schema that validates but doesn't connect is one of six symptoms covered in the full map at how do I tell if my website is losing me clients.