Structured data names things. It does not rank them, crawl them faster, or make them trusted.

Structured data is machine-readable description of meaning on a page: types, properties and relations. In search work the usual vocabulary is Schema.org and the usual syntax is JSON-LD, written in a script block of type application/ld+json.

Schema markup is structured data that uses Schema.org. JSON-LD is one way to write it. The three words are not interchangeable.

Two jobs. First: help Google understand the content of the page and gather information about objects mentioned on it. Second: make the page eligible for supported rich results when the type is current and the required properties are present.

Google’s introduction page still cites two publisher measurements: Rotten Tomatoes added structured data to 100,000 unique pages and reported a 25 percent higher click-through rate on pages enhanced with structured data compared with pages without it. Food Network converted 80 percent of pages to enable search features and reported a 35 percent increase in visits. Those figures measure enhanced pages, not markup in the abstract.

John Mueller restated in 2025 what he said in 2018: structured data will not make a site rank better. It is used for the search features listed in the developer documentation. A structured-data manual action removes rich-result eligibility. Google states that this action does not, by itself, change web ranking.

In June 2025 Google announced it would phase out several lesser-used structured-data displays. The blog post said rankings were unaffected.

• Arrival. A rendering validator can report a graph that never left with the HTTP response.
• The difference between eligibility and display.
• The difference between understanding a type and inserting a Knowledge Graph node.
• That several popular promises – faster indexing, crawl-budget savings, voice-search wins, authority – are not in Google’s structured-data documentation.

Then a graph written by a script after the response is a graph those systems never see. A December 2024 Vercel/MERJ measurement of crawler traffic reported no JavaScript execution by GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot, Bytespider and Meta-ExternalAgent. They fetched JavaScript files. They did not run them. Common Crawl does not render either. No public re-measurement contradicting that finding was located as of 21 September 2026.

That sentence is about arrival. It says nothing about whether a claim made for markup holds. It does not prove citation.

• Google’s Rich Results Test for eligibility.
• Search Console enhancement reports for whether a supported display is actually being served.
• The schema delivery check: cache-bypassed HTML versus the document after load. Count the two states apart. Do not add them.