Measure a repeated phrase in two versions of an article
Admin ·
A phrase can appear just as often after an edit while its percentage changes because the text is longer. Compare the match count and the denominator together to find the sections that explain the difference.
Compare blue river across two sample drafts
Select Load example and Measure density. Each draft has two paragraphs. For blue river, the first has three matches out of eight two-word windows, or 37.5%. The second has one match out of seven windows, or 14.2857%.
Choose blue river under Target by section in either report. The first paragraph of the first draft has four words. The three two-word windows are blue river, river blue and blue river, so two match.
| Phrase | First draft | Second draft |
|---|---|---|
| blue river | 3 matches out of 8 | 1 match out of 7 |
| river | 4 matches out of 10 | 2 matches out of 9 |
Measure HTML heading sections from saved source
Choose Saved HTML and supply the page source as a file or paste it into First draft. Each heading creates a section, and its own text contributes to the counts. Inline elements inside a sentence keep their text in the same sequence. Block elements and sentence boundaries separate the windows.
The tool reads the parsed body and skips scripts, style sheets, hidden attributes and several embedded elements. It reads anchor text without requesting the destination. Styles applied by CSS and content inserted by JavaScript can make the rendered page differ from the saved source.
WHATWG HTML Living Standard — Parsing HTML DocumentsUnicode Standard Annex #29 — Unicode Text Segmentation
Check the denominator in the JSON report
Download keyword-density.json and verify matches and windows for the same index in targets, phrases and each section’s phrases. Each density percentage equals matches divided by windows times 100. A null density means there was no eligible window of that length.
If the percentage moves after editing, compare both numerator and denominator to see whether the phrase count changed. Search provider guidance addresses repetitive use in context, and the tool makes no ranking inference from a percentage.
Common questions
Are accents ignored when matching keywords?
Canonical forms of the same accent are equivalent after Unicode NFC and lowercase. Cafe without the accent and café with the accent remain different terms.
Can a phrase match across two sentences?
No. The tool measures within one segmented sentence, and in plain text it also keeps each line separate. Browser language data can affect its sentence and word boundaries.
Does the checker infer keyword overlap across a website?
It compares the one or two drafts you provide and only the phrases you enter. Search results, site-wide coverage and page performance require separate evidence.