Word Frequency Analysis: Overused Words and Keyword Density

Every writer has a crutch word. Mine, for a long time, was “actually” — I once found it eleven times in a 900-word article and had no memory of typing it even once. You don’t notice your own repetition while writing because your brain autocompletes right past it. A frequency count notices for you.

The Word Frequency Counter takes any text you paste and returns a table of every word, sorted by how many times it appears, with a percentage of the total. That’s the whole tool. What you do with the table depends on why you opened it.

Editing your own writing

Sort a 1,500-word draft by frequency and the top of the list is predictable: the, a, to, of. Skip those. The interesting part starts around position ten, where content words show up. If “solution” appears 14 times in a product page, your reader felt every one of them.

A rough rule I use: any content word above 1% of the total deserves a look. That’s one occurrence per hundred words — often fine, sometimes a tic. The percentage column in the tool makes this a five-second check.

Set the minimum word length to 4 or 5 to hide most function words, or tick Hide common words to drop about 250 of them outright — articles, prepositions, auxiliaries, pronouns. One thing to know when you do: the percentage column then measures each word’s share of the remaining meaningful words, not of the whole text, so the numbers roughly double. That is the right denominator for comparing two drafts’ vocabulary and the wrong one for checking a density target.

Checking keyword density for SEO

Keyword density — the share of your text taken up by a target keyword — used to be gamed hard. In 2005 people wrote pages where the keyword was 8% of the text, and it ranked. Google’s Panda update killed that in 2011, and today stuffing hurts more than it helps.

Density still matters as a sanity check, though, in both directions:

• If your target keyword sits at 4% or more, the page probably reads like it was written for a robot. Rewrite.
• If it appears once in 2,000 words, search engines may not connect the page to the query at all.

Most SEO writers aim somewhere between 0.5% and 2%. There’s no magic number, but the counter tells you where you stand instead of leaving you to guess.

Where single words stop being enough

A frequency table is a vocabulary report. It tells you which words you reach for; it does not tell you what the page claims. Real keywords are usually two or three words long, and their parts mean nothing apart — “frequency” and “counter” both sitting near the top of the table is no evidence at all that the phrase “frequency counter” appears anywhere in the text.

For that you need n-grams, and a different page: the Keyword Density Checker counts two- and three-word phrases and grades one focus keyword against the natural range. The order that works is frequency first — to see the vocabulary and the repetitions — then phrase density, to check the keyword itself.

What a normal frequency curve looks like

Word frequencies follow Zipf’s law: rank the words and each one appears roughly in inverse proportion to its position, so the second is about half as common as the first, the third about a third as common. In the Brown Corpus — a million words of edited American English — “the” takes 6.9% of all words, “of” 3.6%, “and” 2.8%, and the top ten function words together take about a quarter of the text.

That is the yardstick for the 3% figure everyone quotes as the stuffing threshold. A content word at 3% is not slightly heavy; it is sitting between “and” and “of” in a curve built out of articles and prepositions. Anything a reader would call a real word does not belong there.

Analysing texts you didn’t write

Frequency analysis is older than SEO by a few centuries. Scholars have used it to argue about who wrote which Shakespeare play, and stylometry — attribution by word-usage patterns — helped unmask J.K. Rowling as the author behind the pseudonym Robert Galbraith in 2013. Function-word frequencies turn out to be a fingerprint: nobody uses “upon” at quite the same rate as anyone else.

You probably aren’t unmasking novelists, but the same idea works at smaller scale. Students use frequency tables to study vocabulary in a foreign-language text before reading it. Teachers run essays through a counter to see whether a student’s vocabulary suddenly changed between assignments. Researchers use it as the first step of content analysis: count first, interpret after.

What the counter treats as a word

Any run of letters or digits counts, so “don’t” and “well-known” each register as one word rather than fragments. Punctuation is stripped. Cyrillic, accented characters, and CJK text all work, because the tool splits on Unicode categories rather than assuming English.

That is a slightly different rule from the plain Word Counter, which splits on whitespace alone. The totals agree on ordinary prose but diverge on text full of symbols — a standalone dash is a word there and nothing here. If you only need the length of a draft, that tool is the faster answer and shows its tokenisation rules explicitly.

Case sensitivity is off by default — “The” and “the” merge — but you can flip it on when capitalization matters, like counting proper nouns separately from common ones.

One thing the counter deliberately does not do is stem or lemmatise. “Run”, “runs” and “running” stay three separate rows. Stemming is language-specific, gets English irregulars wrong often enough to mislead, and turns arithmetic you can check by hand into a black box. Raise the minimum length and read related forms as a group instead.

Everything runs in your browser. The text never touches a server, so pasting an unpublished manuscript or a confidential report is safe. If you need the table somewhere else, Download CSV writes every row — not just the 200 shown on screen — as UTF-8 with a byte-order mark, which is what stops Excel mangling accented and Cyrillic words.

Paste something into the Word Frequency Counter and check the top of your own list — you already have a crutch word, you just haven’t met it yet.