How the comparison works
Both sides are parsed as JSON, their object keys are sorted recursively, and each is written back out with the same indentation. The two normalised documents are then compared line by line and word by word. Two files that hold the same data therefore compare as identical, however they were formatted or ordered.
Arrays keep their order, because the order of items in an array is part of the data: [1, 2] and [2, 1] are reported as different. Both sides must be valid JSON. If one is not, the page says which side failed to parse.
When to use it
Typical cases are comparing two API responses (before and after a deploy), checking whether a config file differs between environments, reviewing a change to a fixture or snapshot file, and confirming that a re-serialised export still holds the same data.
In the diff view you can switch between split (two columns) and inline (one column), and show only the changed lines. To compare YAML or XML the same way, use Diff Checker and pick the YAML or XML mode.