Density is a symptom, not a lever
Writing to a density target produces text that reads like it was written to a density target, which is exactly the thing search engines got good at detecting. The useful direction is the other way round: write the page, then look at what the counts say it is about, and check that matches what you meant.
Two findings are worth acting on:
- The subject is missing. The top phrases are navigation, boilerplate or the brand name. The page has not actually said what it is about.
- One phrase dominates. Above about 5% for a specific phrase, the text usually reads badly to a person. That is reason enough to fix it.
Stop words, and why single words mislead
Excluded by default, because "the" being your most frequent word is not a finding. But a phrase containing a stop word often is the subject — "cost of living", "return on investment" — so those are kept when looking at pairs and triples.
Paste the copy, not the page
Include the body text and headings. Leave out navigation, the footer and the cookie banner: they appear on every page, so they will dominate the counts and tell you about your template rather than your content.
Questions
What is the ideal keyword density?+
There isn't one. Density was a ranking factor in the early 2000s and has not been one for well over a decade — search engines model meaning, not repetition. Any figure quoted as a target, whether it is 1% or 3%, is folklore.
Then what is this useful for?+
Diagnosis, not optimisation. It shows what a text is about when read mechanically, which is useful for spotting that a page meant to be about mortgage rates actually talks mostly about the bank, or that a translated page lost its subject entirely.
Why does it measure phrases and not just words?+
Because meaning lives in phrases. "Cost", "of" and "living" tell you nothing individually; "cost of living" tells you what the page is. Single-word density is the least informative of the three views here.
How is density calculated?+
As the share of all words the phrase occupies. A two-word phrase appearing ten times in a hundred words is 20%, because it takes up twenty of them. Reporting that as 10% — the phrase count over the word count — understates how much of the text it is, and is what most tools do.