How to compare repeated words before and after an edit

This guide uses a small illustrative example, not a customer draft or an AI-generated rewrite. It explains the calculation, then shows how to apply the same comparison to your own writing.

Start with two versions of the same passage

Keep a copy of the original before editing. For a useful comparison, select the same passage in both versions: a paragraph against its revision, or a whole draft against the revised draft. Including a heading in only one version changes the total and can make the comparison harder to interpret.

1. Open the word frequency counter and turn on Compare an edited version.
2. Paste the earlier version into Original text and the revision into Edited text.
3. Choose whether case matters. Leave the ignore list empty for your first comparison.
4. Select Compare text, then find the word you want to review. Comparison rows are alphabetical.

You do not need to rewrite anything automatically. Counting happens locally in your browser, and neither draft is sent away for counting. Changing an input or option clears the old result; run the comparison again to see the new counts.

A worked example: fewer occurrences, a smaller share

Use clear clear prose as the original and clear simple prose as the revision. Each version contains three counted words. The word “clear” appears twice before the edit and once afterward.

The formula is occurrences divided by all counted words, multiplied by 100. Here, 2 ÷ 3 × 100 gives 66.67%, while 1 ÷ 3 × 100 gives 33.33%, rounded to two decimals.

The count change is −1. The share falls by 33.33 percentage points, calculated from the unrounded values. That is different from saying it fell by 33.33 percent: the occurrence count fell by half. Subtracting the displayed, already-rounded percentages can give a slightly different answer.

Read counts and percentages together

A lower count does not always mean a lower share. Suppose a word appears four times in a 100-word draft, then three times in a 50-word revision. Its count drops from four to three, but its share rises from 4% to 6%: an increase of two percentage points.

That second example is arithmetic, not a measured editing outcome. It shows why each draft needs its own denominator. Check the total word counts before deciding that a revision has become less repetitive. Words added or removed elsewhere can change the percentage even when the word you are checking stays untouched.

Keep the counting rules consistent

The counter recognizes Latin-script words, including accented letters. It is useful for English and other Latin-alphabet text, but it does not count Chinese characters or other non-Latin scripts as words. Numbers and punctuation are excluded. For example, “draft 2026” contributes one word; “version2” contributes the letter sequence “version.”

Internal apostrophes stay within a word, so “don’t” is one word. Straight and curly apostrophes are treated alike. Hyphens separate words, so “well-written” contributes two. Different forms such as “edit” and “editing” remain separate; the tool does not group meanings or word families.

By default, “Clear” and “clear” share a count. Turn on Case-sensitive counts to keep them separate. The selected setting applies to both drafts. Use the same setting across comparisons if you want the results to be comparable.

Hide distracting words without changing the denominator

Enter words in Ignore words, separated by commas or new lines, to hide their rows. Ignored words still count toward each draft’s total. In the three-word example, ignoring “prose” hides that row but leaves the original share of “clear” at 2 out of 3, not 2 out of 2.

This makes it easier to focus on a few terms while keeping the percentages tied to the full passage. The ignore list follows the same case setting. It is a display filter, not a way to remove words from your draft.

Save the comparison, then review the sentences

Use Copy table for a tab-separated table or Download CSV for the full set of result rows. Comparison exports include before and after counts, count change, both percentages, and percentage-point change. Percentages are rounded to two decimals; hidden words stay out of the exported rows.

Each text field and the ignore field allow up to 8,000 characters. Over-limit input is rejected rather than silently shortened. For longer drafts, compare matching sections and keep their totals separate; averaging section percentages does not give the whole-document percentage.

Finally, read the sentences around the repeated word. Repetition can be deliberate, and replacing an important term with a synonym can make a passage less precise. A frequency table measures repetition, not writing quality or whether text was written by AI. Compare your two drafts, then use the counts to decide where a closer read would help.