Line mode for structured material
Use lines for code, configuration, lists, and paragraph-oriented drafts. Inserted blocks remain easy to scan.
Word mode for edited prose
Word comparison highlights rewritten phrases without marking an entire paragraph as replaced.
Character mode for exact strings
Character mode helps with identifiers and short values, but long documents can produce a noisy and expensive result.
Normalize only with intent
Ignoring case or whitespace can hide formatting changes that matter. Record the options with the exported result.
Choose granularity with a short trial
Start with line mode when the sources contain paragraphs, records, source code, or configuration because it preserves the surrounding structure. If a changed line is too coarse, repeat only that passage in word mode. Reserve character mode for short identifiers, punctuation, or spelling changes: its token matrix grows with both inputs, so the tool deliberately stops oversized character comparisons instead of freezing the tab. This staged workflow usually produces a result that a reviewer can scan and still trace back to the originals.
Read additions and deletions as evidence, not meaning
The exported notation prefixes removed lines with a minus and inserted lines with a plus; word and character modes wrap deletions in square markers and additions in curly markers. Unchanged tokens support the similarity figure, but that percentage is not plagiarism, authorship, or semantic similarity. Save the selected mode and normalization settings with any review record, then inspect consequential changes in the source application where formatting, comments, and document metadata are still available.