A list of single words answers the wrong question
Nobody repeats "risk". They repeat "risk assessment". Count words on their own and the phrase that is actually doing the repeating disappears, split across two unremarkable rows, while the top of the list fills with "the" and "of".
So this counts phrases too: every two and three word run in the text. That is the list you can act on, and it is the list an SEO check actually needs, because no search engine has ever cared how often you wrote "the".
"Manage" and "managing" are the same habit
Counted separately they are three or four unremarkable rows. Counted together they are one obvious one. That needs the word endings stripping, which is a rough instrument: strip too hard and "university" and "universe" end up in the same row.
So it only removes the endings that genuinely mean the same word in a different place in a sentence, it is a switch rather than the default, and every row lists the forms that went into it. A total you cannot check is a total you should not trust.
On keyword density, briefly
There is no target percentage. There never was an official one, and search engines stopped rewarding density well over a decade ago. What they do now is treat conspicuous repetition as a signal that a page was written for a machine.
The useful threshold is a human one: past about 4% of the text, a reader starts to notice. If a term is above that, the fix is not to delete instances until the number comes down. It is to find the sentences that would read better with a pronoun, or with nothing at all, which usually turns out to be most of them.
How much text you need
A few hundred words at minimum. In a paragraph, one word used twice is already top of the list, and that is arithmetic rather than a finding. The tool says so rather than letting you draw a conclusion from six sentences.
The variety number
Distinct words over total words. Ordinary prose lands around 40 to 50%. Below roughly 30% a small handful of words are carrying the whole piece, which is almost always several sentences saying the same thing in the same shape rather than any individual word being overused.
That one is worth reading the piece back for, rather than editing against. It is the one measure here that points at structure instead of vocabulary.
Everything runs on your own machine. Nothing is uploaded and nothing is stored, so an unpublished draft is safe to paste.
Common questions
Why is the top of my list full of "the" and "of"?
Because they are the commonest words in every piece of English ever written, so they tell you nothing about yours. Leave common words out, which is the default here, and the list becomes about your subject rather than about the language.
The single word counts do not show the repetition I can hear.
They usually do not, because people repeat phrases rather than words. Nobody repeats "risk", they repeat "risk assessment", and counting single words splits that across two unremarkable rows. Switch to two or three word phrases and the thing you can hear will normally be at the top.
What does "count word endings together" do?
It puts "manage", "manages", "managing" and "managed" in one row rather than four. Four rows of one look like nothing; one row of four looks like a habit. The endings are stripped roughly rather than by dictionary, so each row lists the forms that went into it and you can check the total.
What is a reasonable percentage for a keyword?
There is no official number and anyone quoting one is guessing. As a rule of thumb, past about 4% of the text a person starts to notice, which is the only threshold that matters. Search engines stopped rewarding density many years ago and now treat obvious repetition as a negative signal, so writing to a target is worse than not thinking about it at all.
What does the "words that are different" percentage mean?
Distinct words divided by total words. Ordinary prose sits somewhere around 40 to 50%. Much below 30% usually means several sentences are saying the same thing in the same way, which is worth reading back rather than editing word by word.
How much text do I need for this to mean anything?
A few hundred words at least. In a short passage a word used twice is already at the top of the list, and that says nothing about the writing. The tool tells you when there is not enough to conclude anything.
Does it count "don't" as one word or two?
One. Apostrophes are kept inside a word, including the curly apostrophe that text pasted from Word carries, which otherwise turns every contraction into two fragments and quietly inflates the count.
Does the text I paste go anywhere?
No. It runs in your browser on your own machine, with no server and no account, and nothing is stored. An unpublished draft is safe to paste.