Use text statistics to edit, not to manufacture quality.
Start with a reader decision
Suppose a repair guide repeats “check the cable” in several steps. A high frequency may be necessary because each step stands alone, or it may reveal an unclear sequence. A counter cannot choose between those explanations. Write down the question the article answers before looking at a table of words.
Make comparisons reproducible
Paste only the article text, excluding navigation and the footer. Select the same language for both drafts. This tool uses locale-aware word segmentation, retains common words and reports occurrences over all word tokens. Chinese, Japanese, Korean and mixed-language passages do not have the same word boundaries as English; record the language alongside your results.
Interpret numbers narrowly
If a word occurs 12 times in 600 word tokens, its share is 2 percent. That arithmetic says nothing about whether 12 mentions are appropriate. There is no target density in this workflow. A technical definition, checklist or quotation can produce a concentrated vocabulary without being manipulative. Read the actual sentences before removing a repeated term.
Edit for missing decisions
Look for the moment where a reader must choose what to do next. Does the guide explain what a failed check means? Does it distinguish a loose cable from damaged insulation? Add the missing explanation rather than padding the article to reach a word count. In this example, any safety-sensitive instruction would also need a qualified technical review before publication.
Check substance after the rewrite
Run the statistics again with the same input boundaries and language, then read the text aloud or in sequence. Keep changes that improve clarity even if the word count rises. Ask another reviewer to verify factual claims and ambiguous terms. These tools describe the text you provide; they cannot establish originality, expertise, accuracy or approval by an advertising network.