Check keyword density and word frequency

What keyword density actually measures

Keyword density is occurrences divided by total words: use a word 12 times in a 480-word page and its density is 2.5%. That is all it is — a ratio, not a ranking signal you can tune. The famous advice to aim for 2–3% dates from a time when search engines really did score pages partly on raw term frequency, and it has survived two decades past its expiry. Modern retrieval models topics, synonyms, entities, and intent; there is no published evidence that hitting any density figure improves rankings, and mechanically repeating a phrase to reach one produces exactly the stuffed, unreadable page that quality systems demote.

What a frequency table is genuinely good for is diagnosis. It catches keyword stuffing before a search engine does — if one content word carries 4% of a page, a human reader has already noticed the repetition. It shows whether a page about a topic actually uses the words of that topic, or dances around them. And it exposes the tics no author sees in their own draft: the paragraph opener used nine times, the crutch adverb in every second sentence, the metaphor that felt fresh the first three times it appeared.

How to read the three tables

The single-word table answers "what is this text about, and what does it lean on?" With the common-words filter on, function words — the, and, of, to — are hidden, because on their own they dominate every English text and say nothing about yours. Turn the filter off when the function words themselves are the question: an over-reliance on "very", "just", or "really" is a style finding, not noise. The minimum-length control does the same job more bluntly, hiding everything shorter than the cutoff.

The phrase tables answer a different question — "what do I actually say repeatedly?" — so the filters deliberately leave them alone. Function words inside a phrase are load-bearing: "how to", "out of stock", "terms of service" are the units people write and search, and filtering them would leave fragments that appear nowhere in your text. Phrases are counted within sentences only; a window never crosses a full stop, comma, colon, semicolon, bracket, dash, or line break, because a "phrase" spanning the end of one sentence and the start of the next was never read as one. Every percentage in all three tables shares the same denominator — the total word count — so a phrase's density is directly comparable to a word's.

Using frequency data while editing

For an SEO review, check three things in order. First, that the page uses its target phrase at all — the 2-word and 3-word tables show whether "keyword density checker" appears as a phrase or only as scattered words. Second, that nothing is stuffed: a content word far above every other content word, or a target phrase repeated in near-identical sentences, reads as manipulation to people and machines alike. Third, that variants exist — a page that repeats one exact phrase while never using a natural synonym usually reads like it was written backwards from a keyword.

For prose editing, the tables are a mirror. Sort your draft's top words with the filter on and you will find the noun you have leaned on so hard it has stopped meaning anything; flip the filter off and the crutch words surface. The 3-word table is the most brutal of the three: any trigram that appears three or more times in a short piece is a sentence pattern you are recycling. For plain totals — words, characters, sentences, reading time — the word counter on the homepage runs live on the same text, using the same definition of a word, so the two pages never disagree about your count.

Common questions

What is a good keyword density?

There isn't one, and any tool promising a magic 2–3% is selling advice from the year 2003. Early search engines ranked pages partly on raw term frequency, so a target percentage briefly made sense; modern engines model topics, synonyms, and intent, and no published evidence shows any density figure improving rankings. Frequency is still worth checking — but as a symptom detector for stuffing and repetitive writing, not as a dial to tune upward.

How is density calculated here?

Density is occurrences divided by the total word count of the whole text, shown as a percentage. The denominator always includes every word — even ones the common-word filter hides from the table — so hiding 'the' never inflates the apparent density of everything else. Phrases use the same denominator, which is why a phrase's percentage is directly comparable to a single word's.

Why do the phrase tables still show common words when the filter is on?

The filter deliberately applies to single words only. On its own, a word like 'of' or 'how' tells you nothing, which is why it is worth hiding; inside a phrase those same words are load-bearing — 'how to', 'out of stock', 'terms of service' are exactly the phrases people write and search for. Stripping them would leave mangled fragments like 'terms service' that appear nowhere in your text.

Do phrases count across sentence boundaries?

No. The last word of one sentence followed by the first word of the next was never written or read as a phrase, so counting that pair would only manufacture noise. Phrase windows stop at sentence-ending punctuation, commas, colons, semicolons, brackets, dashes, and line breaks; hyphens and apostrophes stay inside words, so 'well-known' and "don't" survive intact.

Does capitalisation or an accent change the count?

Counting is case-insensitive, so 'Marketing', 'marketing', and 'MARKETING' merge into one entry. Accents are preserved rather than stripped, because 'café' and 'cafe' are different words and folding them together would miscount text that uses both. The two Unicode ways of writing an accented letter are normalised first, so they never split one word into two entries.

Is my text uploaded anywhere?

No. Every count runs in your browser as you type — nothing is sent to a server, logged, or stored, and the page keeps working offline. You can paste an unpublished manuscript or a client's draft without it leaving your machine.