Keyword density is not a thing.

Where the number came from

Very early search engines really did lean on term frequency: a page that used a word more often looked more about that word, so people repeated the word. That era ended a long time ago, and it ended badly, with keyword stuffing so severe that fighting it was one of the first big jobs the algorithms took on. The two to three percent figure is a fossil from that time, kept alive by tools and templates rather than by anything Google publishes or does now.

Modern search does not read a page as a bag of words to be counted. It works with meaning, with the relationships between concepts, with whether the page actually answers the intent behind the query. In that world "how many times did you say the phrase" is close to irrelevant, and a page can rank perfectly for a term it uses once, or not at all in the exact form, because the page is clearly the best answer.

Why a density score is a rule the tool made up

When an audit tool tells you the density is too low and flags it red, ask where the target came from. It came from the tool. Google has never published an ideal keyword density, because there is no single figure that could be right across a legal disclaimer, a recipe, a product page and an essay. The tool picks a band that sounds reasonable and grades every page against it, which produces a number that looks objective and means nothing.

This is the same trap as every proxy that becomes a target. The instant "hit 2.5 percent" is the brief, the cheapest way to satisfy it is to jam the phrase in until the meter turns green, and you have optimised the meter rather than the page. The writing gets stiffer, the repetition gets obvious, and if you push it far enough you arrive back at the stuffing the whole system was built to punish.

What actually matters on the page

The useful version of the old instinct is real and worth keeping: the page should obviously be about its subject. That means using the words a reader and a search engine would expect to find, the related terms and the sub-topics, the questions around it, the specifics that only someone who knows the subject would include. That is coverage and depth, not frequency, and it is the substance of good SEO copywriting: answering the query fully in language a person actually uses.

It also means the mechanical on-page basics that do carry weight, none of which is a percentage: a title and headings that name the subject clearly, a first paragraph that gets to the point, internal links that place the page in context. Those are on-page decisions about clarity and structure, and they do far more than any amount of fiddling with how many times the phrase repeats. The same instinct sends people to write a separate page for every phrasing, which is the same mistake at the level of the site rather than the paragraph.

The general lesson

A number that is easy to compute is not the same as a number that means something, and keyword density is the purest example in the whole field: precise, measurable, gradeable, and disconnected from how search actually works. Write the page for the person who needs it, cover the subject the way someone who knows it would, and delete the density meter from your process. It is measuring a rule that stopped existing before most of the sites using it were built.