Compare JSON: the JSON Diff tool for exact changes

How to compare JSON documents

To compare two JSON documents, paste JSON into Original JSON and Changed JSON, set the comparison options, then select Compare JSON. The Results table lists every path that changed.

1. Paste JSON into Original JSON and Changed JSON, or select Choose JSON file on either side to read a UTF-8 file instead; the filename or extension does not matter.
2. Ignore object key order is on by default, so a reordered but equal object counts as no change. Turn it off to report key order too. Then choose Number comparison: Exact literals (the default) or Numeric values.
3. Select Compare JSON. Cancel stops a long comparison; Load example fills both boxes with a short sample pair.
4. Use Show to filter the Results table to All changes, Added, Removed or Changed, and Previous or Next to page through a long list. Copy report or Download report saves the complete comparison, including rows not shown on the current page.

Comparison rules

Numbers stay exact by default: 9007199254740993 is reported as changed against 9007199254740992, even though both round to the same value in ordinary floating-point math. Exact literals mode also distinguishes 1, 1.0 and 1e0 as different written values; Numeric values mode compares their decimal value instead, so those three compare equal to each other.

Array elements are compared by index, not by matching content, so inserting a value shifts every later index: the comparison reports each later index as changed, and the last one as added, even though only one value is genuinely new; the tool does not detect moves. Adding an object or array reports the container itself as added, together with each of its descendant paths. A key containing a slash or a tilde is escaped in its JSON Pointer path, as the format requires: a tilde becomes ~0 and a slash becomes ~1, so a key literally named a/b appears in a path as a~1b, not as two path segments.

What counts as equal

Formatting differences never produce a change: pretty-printed and minified versions of the same document compare equal, and so do 2 documents with different indentation or line endings. Equivalent string escapes are also ignored, so a string written as "\u0041" and one written as "A" compare equal, since both decode to the same text. A single leading UTF-8 byte-order mark on either document is ignored as well, rather than treated as part of the content.

A duplicate key within one object is rejected outright, on either side, rather than silently keeping the last occurrence the way some JSON parsers do. JSON itself does not define which repeated key should win, and a JSON Pointer path has no way to address one occurrence over another, so the comparison reports a duplicate-key error naming the affected side instead of guessing.

Limits

This tool sets no size or depth limit. Your browser sets two: its memory, and its call-stack depth for deeply nested input. JSON nested several thousand levels deep can stop with the message that the browser could not finish, and a file too large to read as text stops with a text-size error. A very large or very different pair of documents can also simply take longer to compare than two small, similar ones, and Cancel stops the comparison at any point; comparing a smaller section of a large document is the fallback if a full comparison does not finish.

The Results table shows 50 changes at a time; use Previous and Next to page through the rest. The count of Added, Removed and Changed rows above the table always reflects the complete result, not just the visible page.