Structured data is a promise made in a language nobody reads by hand. A single missing comma turns the whole block into nothing; a relative image URL turns it into a promise that cannot be kept. This check pulls every ld+json block out of the served page and reads the nodes one by one, against the fields search engines expect for each type.
From this server: exactly one request for the page, capped at 1.5 MB. Nothing is executed, no cookies are sent, nothing is stored beyond the rate counter.
Limits worth knowing: only JSON-LD in the served HTML is read. Microdata and RDFa are not — they sit scattered through the markup and would be a second tool. Blocks a script inserts later stay invisible, and required fields come from what search engines ask for today, which changes.