This finds the longest common subsequence of lines between two texts — the same class of algorithm behind tools like diff — so it can tell apart lines that are genuinely unchanged from ones that were added, removed, or modified. Lines that were changed rather than purely added or removed get a second pass of word-level diffing so you can see exactly which words shifted inside an otherwise-similar line, and the whole comparison can be viewed side-by-side or as a unified patch-style list.
They normalize each line before comparing it — trimming and collapsing repeated spaces for whitespace, lowercasing everything for case — so two lines that differ only in indentation or capitalization show up as unchanged. The original text in the textareas is untouched; normalization only affects how the comparison decides what counts as "the same."
Side-by-side shows the original and changed text in two columns, lined up so you can scan across a row to see what replaced what — good for spotting structural differences. Unified view interleaves both versions into a single column with +/− markers per line, closer to a traditional patch format, which is easier to read top-to-bottom when changes are sparse.
It copies a plain-text version of the diff using leading +, -, and blank markers per line — similar in spirit to a unified diff format — so you can paste the comparison into a chat, ticket, or document without needing a rendered table.