A ranking factor that stopped being one
Keyword density mattered when search engines counted words. Repeat a phrase often enough and you ranked for it, which is exactly why the tactic died — it was trivially gamed, and the pages that won were the least readable ones.
Modern ranking reads a page for topic and meaning. It recognises that car insurance, vehicle cover and motor policy are the same subject, and no percentage target survives contact with that. The frequently quoted ideal of one to two per cent has no basis in anything current.
So the measurement is worth having, but as a check on your writing rather than a score to optimise. The Keyword Density Checker reports the eight most frequent words in your text and their share of the total, which tells you what your page appears to be about.
What the count is actually doing
Your text is lowercased, stripped of punctuation and split on whitespace. Words of three characters or more are counted, and the top eight are reported with their percentage.
Two consequences worth knowing. Short words are excluded, so anything under three characters never appears. And a target keyword is matched as a whole single word, which means entering a two-word phrase returns zero — not a bug, just the limit of word counting.
That single-word limitation points at the real weakness of density as a concept: most search queries are phrases, and counting individual words tells you very little about whether your page addresses a phrase.
Read the list, not the percentages
The useful output is which words appear, not how often. Scan the top eight and ask one question: does this list describe the page I meant to write?
- The topic is missing entirely. A page about pricing whose top words are features, platform and solution is not about pricing.
- Filler outranks substance. If also, however and really are placing highly, the writing is padded rather than the keywords being wrong.
- Your brand name dominates. Common on service pages and a sign the page describes you rather than the reader’s problem.
- One word is far ahead of everything else. Usually repetition you did not notice, and it reads worse than it counts.
Over-optimisation still has a cost
Density is not a ranking factor, but stuffing is still a problem — just for different reasons than people expect.
Search guidance treats text written for engines rather than readers as spam, and unnaturally repeated phrases are the clearest example of it. That risk is real, though it applies to obvious abuse rather than to a page that mentions its subject a few times.
The larger cost is human. Repetition reads badly, and a reader who finds the same phrase in four consecutive sentences leaves. Engagement is measurable in a way density never was, and losing readers is a more direct route to losing rankings than any percentage.
What to measure instead
- Whether the page answers the query. Read the top-ranking results for your term. If they answer something yours does not, that gap matters more than any word count.
- Coverage of the subject. A genuine article on a topic naturally contains the related terms — a page about mortgages that never mentions interest, term or deposit is thin regardless of how often it says mortgage.
- Headings and first paragraph. Placement carries far more weight than frequency. The subject stated clearly in the opening does more than twenty mentions further down.
- Click-through from Search Console. Impressions without clicks means your title and description are failing, which no density figure will reveal.
- Whether a reader stays. The one signal that survives every algorithm change.
A reasonable use for the number
Compare, do not target. Run your page and run the page currently outranking you. If theirs mentions the subject naturally throughout and yours mentions it twice in an aside, you have learned something — not a percentage to hit, but that your page is less about its subject than you thought.
It is also a fast check on a generated draft. Model output tends to circle a phrase, and seeing one word at an implausible share is a quicker signal than rereading twelve hundred words.
Beyond that, write for the person who searched. Mention the subject where it belongs, use the words a real person would use, and let the count be a diagnostic rather than a goal.
The analysis runs entirely in your browser. Nothing you paste leaves the page or is stored here.