Why the percentage is not a target
Keyword density had a brief technical justification in the late 1990s, when ranking was driven largely by term frequency within a document. That has not described how search works for a very long time. Modern retrieval weighs terms against how common they are across the whole index, considers the query's meaning rather than its literal string, and reads the surrounding passage. Repetition rate is not an input you can turn up.
The stronger point is that Google's own spam policies name keyword stuffing directly, describing it as loading a page with repeated words to manipulate ranking, and treating it as a violation rather than a technique. So the percentage is not neutral advice being applied cautiously. Chasing it is either useless or actively harmful, with no version where it helps.
Ranges like "one to three percent" circulate widely and have no published source behind them. They came out of early optimisation folklore and survived by repetition. If you must think in numbers, the honest framing is descriptive: pieces genuinely about a subject tend to land under three percent for their main term without the author counting. That is what naturally-written text does, not a level to write toward.
What counting is genuinely good for
Three things, none of them about search engines.
First, catching unintentional repetition. Writers have tics, and a word you leant on eleven times in a thousand is invisible while you are writing it and obvious to a reader. The frequency list surfaces it. Second, confirming a term is there at all. Pieces drift during editing, and it is entirely possible to publish something you believed was about compaction that mentions compaction twice. Third, seeing distribution. A term appearing only in the second quarter usually means the piece raised a subject and moved on without closing it.
All three are editorial observations you would make yourself on a careful read. Counting just makes them faster and less dependent on being in the mood to notice.
How each figure here is worked out
The word count splits on anything that is not a letter, digit or apostrophe. Term counts are literal, case-insensitive substring matches, so "slab" is counted inside "slabs" and inside "slabwork". That is deliberate — it matches how a reader experiences the repetition — but it means a very short term will report inflated counts, which the warnings flag. Share is the count multiplied by the number of words in the term, divided by the total, which is the most common of several conventions; other tools use others, so numbers from different tools will not agree to the decimal place.
The distribution bars split the draft into four equal spans by character position, not by paragraph or section, so a piece with one very long section will not divide where you would divide it by eye.
The failure mode this tool is designed not to have
Most density checkers show a coloured verdict: green in the target band, red outside it. That design decides for you that a band exists, and the number people then write toward is one that nobody can defend. This tool reports counts, placement and distribution, and declines to grade the result, because a grade would be an assertion about search behaviour that neither this page nor any other page is in a position to make. If a tool promises that a percentage will improve your ranking, the interesting question is what evidence it has, and the answer is generally none.
Questions people ask
What keyword density should I aim for?
None. There is no figure to aim for, because repetition rate is not something search engines reward. Text written naturally about a subject usually falls under three percent for its main term, but that is an observation about what natural writing does, not a specification to meet. If you edit a draft to raise or lower a percentage, you have made the writing worse in exchange for nothing. The one number here worth acting on is a term appearing zero times when the piece is supposed to be about it.
My term shows zero but I definitely wrote it.
Matching is literal, so the exact characters have to be present. The usual causes are a hyphen on one side and not the other, a plural where you searched the singular in a way that does not nest, a possessive apostrophe, or a different spelling entirely. Case is ignored, so that is not it. Check the frequency list for the form that actually appears in the draft and use that as your term.
Why does this give a different percentage from another density tool?
Because the convention differs. Some tools divide by total words, some by unique words, some count a multi-word phrase as one occurrence and some as one per word in the phrase, and tokenisers disagree about hyphens, numbers and apostrophes. This tool counts a phrase as occupying its own word length. Differences of half a percentage point between tools are routine and none of them is more correct, which is another reason not to treat the figure as precise.
Can this tell me whether my page will rank?
No, and nothing that only reads your text can. Ranking depends on what else exists on the query, how the site is regarded overall, links, how well the page matches what the searcher wanted, and a great deal more that is not visible from inside a single document. This page counts words in a text box. That is the entire scope of it.