Keyword Density Checker

Density is a diagnostic, not a target. There is no percentage Google rewards, and anyone selling you one is selling you 2009. What the number is good for is the two ends: a page whose subject appears twice in two thousand words is not about that subject as far as a machine can tell, and one where it appears at 6% reads as written for a machine, because it was.

Underneath, it is a word frequency count with n-gram phrases built on top of it — the plainest content analysis there is. The plainness is deliberate. Nothing here infers intent, scores your writing or compares you against a competitor, because every one of those would be a guess dressed as a measurement.

Arabic is why this exists next to a dozen free ones. Splitting on spaces counts أدوات and الأدوات and وأدوات and بالأدوات as four separate terms. The word the page is actually about arrives spread across a dozen rows and never reaches the top of the list — which is the one thing the list is for. The article, the conjunction and the prepositions that glue onto a noun are stripped before counting.

The stemming is deliberately shallow. A real stemmer would fold أدوات to أداة, and would also fold words a reader considers distinct — which makes a report harder to trust than the problem it solved. Only what is unambiguously glued on comes off.

Phrases are measured by the words they occupy. A three-word phrase appearing ten times is thirty words of a thousand-word page, not ten. Reporting it as 1% understates it threefold, which is how a stuffed phrase hides in a clean-looking list.

The sample is Arabic on purpose, and it is the demonstration: five forms of one word, which any space-splitting counter reports as five terms with a count of one each. Here they arrive as one row with a count of five, which is the true shape of the text.

Stop words are excluded from the ends of a phrase, not the middle. "rate of return" is a phrase; "of return" is not. Dropping stop words entirely would break the first one, and keeping them entirely fills the list with "the".

Under 300 words the number stops meaning much. One repetition swings the percentage several points, so the tool says so rather than reporting a precise figure that is mostly noise.

For the structure of the page rather than its vocabulary, the heading checker reads the outline.