How the line comparison works
A diff identifies the smallest useful sequence of edits needed to describe how one version became another. This PagesTools checker compares complete lines. Unchanged lines receive a neutral prefix, lines found only in the original are marked as removed, and lines found only in the revised text are marked as added. The result resembles a compact unified review.
Line-based comparison is especially readable for source snippets, configuration, lists, transcripts, and prose with meaningful paragraph breaks. A changed sentence on one line appears as one removal and one addition rather than as highlighted individual letters. That choice keeps the output predictable and avoids implying that a word-level interpretation is available when the algorithm is comparing lines.
Repeated lines can make more than one alignment mathematically reasonable. The checker chooses a bounded alignment that preserves useful surrounding context, but another diff program may place an addition slightly earlier or later. The substance is the same: inspect the marked content and nearby unchanged lines instead of treating the exact position of a repeated line as semantic proof.
A reliable review workflow
Place the known earlier version in the original field and the candidate version in the changed field. Preserve the line breaks that matter to the document. Run the comparison, then read from the first marked change through its unchanged context. If the output is unexpectedly noisy, check whether an editor reformatted paragraphs or converted line endings before assuming every sentence was rewritten.
For code or configuration, use the diff to locate edits, then validate the revised text with the relevant compiler, parser, or test suite. A diff explains textual change; it does not establish that the new syntax is valid or the behavior is correct. For contracts and policies, use it as a navigation aid and have an appropriately qualified person interpret the effect.
When collaborating, copy or download the result with both source versions still available. A diff alone cannot always reconstruct intent, authorship, or review status. Record the document names, dates, and revision identifiers in your normal version system. That context prevents a technically accurate comparison from being mistaken for an approved or final change record.
Whitespace, formatting, and difficult inputs
Spaces and tabs are part of each line, so indentation changes can produce removed and added entries. This is desirable for whitespace-sensitive code but may distract during a prose review. If formatting is irrelevant, normalize it in a copy before comparing. Do not normalize the only original, because doing so destroys evidence about the exact source text.
A moved block generally appears as removals at its former position and additions at its new position. The checker does not claim to detect moves, plagiarism, paraphrases, or equivalent meaning. Likewise, replacing one word changes the containing line. Use a specialized semantic or word-level review when the question depends on meaning rather than exact line content.
Very large documents require more memory and comparison work. The implementation uses bounded input and may reject material that would make an interactive browser unreliable. Split a large file at stable section boundaries, compare the sections separately, and preserve their order. Binary files, rich document formatting, comments, and tracked-change metadata are outside this text-only tool.
Local privacy and review limitations
Both versions are processed locally in your browser. PagesTools does not need to upload their contents to produce the comparison. Local processing is helpful for drafts and internal notes, but confidentiality still depends on the device, clipboard, browser extensions, screen sharing, and any software that can observe the page. Follow your organization’s handling rules for restricted documents.
The result is not a cryptographic audit, legal redline, authorship record, or proof that no change exists. Text may have been normalized before it reached the fields, and non-text features are not represented. For high-stakes review, compare the authoritative files with approved tooling, verify signatures or hashes where applicable, and retain the original evidence alongside this convenient visual check.