Compare two text versions only after fixing which one is the approved baseline and which one is the candidate. Use line view to locate changed sections, word or character view to inspect details, and then read the complete sentences in the file or message that will actually be sent.
Establish a clean baseline and candidate
A useful comparison begins with two named versions, not two windows that may both contain unsaved edits. Preserve the source files, label the older approved copy as the baseline, and save the proposed copy separately before extracting the same text range from each.
Consider a supplier notice that originally says, “Orders confirmed by 5:00 p.m. ship within seven business days.” The candidate says, “Orders received by 5:00 p.m. ship within seven days.” Those edits change the trigger and the meaning of the deadline even though most words are unchanged. The comparison needs the complete sentence, its heading, and any nearby exception that controls it.
Paste the baseline on the left and the candidate on the right in the Text Diff tool. The page compares the text you paste; it does not open a DOCX, PDF, email attachment, or folder. Decide beforehand whether headers, footnotes, tracked changes, comments, and attachment names are part of the review. If they matter, extract and review them through an appropriate document workflow rather than assuming a plain-text comparison contains them.
Keep sensitive content out when the task does not require it. Replace customer names, account numbers, private prices, and access details with consistent placeholders in both samples. If organizational policy prohibits pasting the material into a browser tool, use an approved local comparison system instead.
Choose line, word, or character view for the question
The three modes answer different questions. Start by line when the text has paragraphs, clauses, release-note bullets, configuration lines, or one record per row. It makes inserted and deleted sections easy to locate without turning every changed letter into a separate distraction.
Switch to word view for English prose after locating a changed sentence. Word mode separates on whitespace, so it can expose phrases such as “business days” becoming “days,” or “may cancel” becoming “will cancel.” It is less helpful when punctuation and spacing carry the important distinction.
Use character view for dates, amounts, identifiers, URLs, punctuation, and short clauses. A single digit in 2026-08-01, a missing minus sign, or a changed URL parameter can matter more than an entire rewritten paragraph. Character mode also reveals typographic apostrophes, doubled spaces, and small punctuation changes that word mode can group awkwardly.
The tool performs exact comparisons. It does not automatically ignore case, normalize line endings, collapse spaces, or decide that two punctuation styles are equivalent. When a formatting conversion produces hundreds of unimportant whitespace changes, do not erase the originals. Make a separately named normalized comparison copy, document what was normalized, and return to the untouched files for final approval.
Review meaning-changing edits before cosmetic ones
Highlighted additions and deletions do not tell you whether an edit is correct. Triage differences by potential impact, then return each important change to its requirement, approval note, or responsible owner.
Review these categories first:
- Negation and obligation:
can,cannot,must,may,only, andunless - Quantities: prices, percentages, units, limits, and version numbers
- Time: dates, time zones, business versus calendar days, and start conditions
- Scope: product names, regions, audiences, exclusions, and attachment references
- Destinations: email addresses, URLs, form names, and support channels
For the supplier notice, the text compare tool can show that “confirmed” became “received” and “business days” became “days.” It cannot know whether the supplier authorized those changes. Mark each substantive difference as accepted, rejected, or awaiting confirmation, and record who owns the decision.
Purely cosmetic changes still deserve a quick scan. A heading may have been demoted, a bullet may have lost its marker, or a link label may no longer describe its destination. However, fix high-impact meaning first so a large punctuation cleanup does not hide an altered deadline.
Compare focused sections instead of one enormous paste
Long material is easier to review in stable sections. Compare the title and summary, then each numbered clause or announcement section, using the same boundaries on both sides. A section-by-section process also makes it obvious when one version omitted an entire block.
Use a simple review note for each section: baseline name, candidate name, comparison date, substantive differences, and unresolved decisions. Do not treat a screenshot of red and green highlights as the only record. The source text, approval evidence, and final delivery copy are what future reviewers will need.
If the material was copied from a PDF or rich document, watch for extraction artifacts such as repeated headers, broken lines, missing symbols, or reordered columns. A clean diff may only prove that two flawed extractions are identical. Open the source documents and visually inspect the affected pages before approving the text.
The browser tool displays the comparison result but does not save a version history or export an approval report. Keep records in the system your team already uses for controlled documents, tickets, or releases.
Verify the actual final copy after editing
The final check belongs after the accepted edits have been applied to the real destination. Reopen the email draft, publishing form, contract copy, or release page and extract the final text again. Compare it against the approved candidate, not against memory.
Use the Word Counter when the destination has a length limit or when a large count change may indicate a missing section. Equal counts do not prove equal content, and different counts do not identify the problem; the count is only a quick signal to investigate.
For Markdown announcements, open the final source in the Markdown Editor to inspect headings, lists, links, and rendered structure. Its preview cannot prove that a production site uses the same renderer, so check the actual staging or delivery environment when layout matters.
Before sending, verify at least one high-risk value from each category that appears: a date, an amount or quantity, a condition, a link, and the named recipient or audience. Confirm that the attachment name and subject line match the approved version. Then run one last text comparison between the approved copy and the text that will be delivered.
Keep the result in proportion to what it proves
An unchanged comparison proves only that the two pasted text values are identical under the selected exact mode. It does not prove that the correct files were copied, that formatting is preserved, that tracked changes are accepted, or that an authorized person approved the content.
When a difference is unexpected, stop the send and trace its source. Do not automatically accept the right side because it is newer, and do not automatically restore the left side because it was previously approved. Resolve the decision, update the named candidate, and repeat the focused comparison.
This disciplined separation—baseline, candidate, decision record, and final delivered copy—turns a text diff from a colorful display into a reliable review step.
Frequently asked questions
Can the text compare tool open two files?
No. It compares two text values pasted into the left and right fields. Open the source files separately and extract the same review scope from each.
Which mode should I use for a contract clause?
Start with line mode to locate changed clauses, then use word mode for English wording and character mode for dates, amounts, punctuation, or identifiers. Read the full clause after reviewing highlights.
Can it compare Word formatting or tracked changes?
No. Plain-text comparison does not include styles, comments, revision history, page layout, headers, or other document structure unless you separately extract meaningful text from them. Review those features in Word or another appropriate document system.
Why does a small edit produce many highlighted differences?
Moved lines, changed whitespace, line-ending conversions, or a rewritten sentence can alter token alignment. Narrow the comparison to the affected section and create a documented normalized copy only when formatting noise is understood.
Does no highlighted change mean the final copy is approved?
No. It means the two pasted values are the same. Approval, source selection, formatting, attachments, recipient choice, and the actual delivered version still require separate verification.