See every field in a JSON document with its types and how often it is present, where the nesting goes, and any duplicate key a parser would silently discard.
Waiting for JSON
Reports the document's shape, where the nesting goes, and which fields are not always present.
The structure appears here
Every field with its types and how often it is present, every mixed array, and any duplicate key the parser would have thrown away.
JSON Structure Inspector runs entirely on your device — nothing is uploaded.
You’re on the Free plan. Compare plans
What shape is this document, where does the nesting go, and which fields are not always there. The questions you have when an API response arrives undocumented and you need to write something that reads it.
The part worth reading is how duplicate keys are found. Parsing a document with a repeated key returns an object with one of the values and no record that anything was dropped — the earlier value is simply gone. Every tool that inspects the parsed object is therefore structurally incapable of reporting a duplicate key, however it describes itself. This scans the source text, which is the only place the evidence still exists.
It matters because the specification permits duplicates and says nothing about which one wins. Two parsers in the same pipeline can legitimately disagree about what a document means, which is a real class of bug and completely invisible from inside a parsed object. Each duplicate is reported with the line of every occurrence and what each one held.
Field presence collapses array indices. The email field of the first record and the email field of the second are reported as one field with a star where the index was, present in some proportion of the elements. A field present in ninety-eight of a hundred records is the thing worth knowing; a hundred separate rows saying the same thing is not. That proportion is what tells you which fields are optional, which is the thing a schema generated from one sample cannot know.
Alongside that: every array whose elements are not all the same kind, which is legal JSON and awkward for anything needing one type per list; every empty object and array, which are positions nothing can be inferred about; and the deepest path in the document, which is usually the answer to why is this response so hard to work with.
By reading the source text rather than the parsed object. Once a document has been parsed the earlier value is gone with no record of it, so the text is the only place the evidence survives. That is why this scans it.
The specification allows them and does not say which should win, so two parsers in one pipeline can disagree about what the document means. It is a real source of bugs and invisible from inside a parsed object.
An array index, collapsed. Every element of a list contributes to one field row, so a field present in most records but not all is reported once with its proportion rather than as a hundred separate entries.
Which fields are optional. A field in ninety-eight of a hundred array elements is almost certainly optional, and that is precisely what a schema inferred from a single sample cannot work out.
Because it is legal JSON that is awkward for everything downstream — a generated type, a typed deserialiser, a database column. Worth knowing before you write code that assumes one type.
No. It is parsed and scanned in your browser, so a response containing real data never leaves the page.