Word Frequency Counter

This counter only understands unaccented English letters, digits and the straight apostrophe

A word, to this tool, is a run of the characters a to z, 0 to 9, and the straight apostrophe, after everything has been lower-cased. Anything else is a separator. That is a deliberately narrow definition and it has consequences you need to know before you trust a count:

• Accented letters are dropped, and the word is counted anyway. Measured: café café reports “2 words, 1 unique” and lists the word as caf. The é is gone, not replaced.
• An accent in the middle splits one word into two. Measured: naïve reports “2 words, 2 unique”, listed as na and ve.
• Non-Latin scripts count as nothing at all. Measured: the Russian for “hello world” reports “0 words, 0 unique”. No error, no warning, just zero.
• Hyphenated compounds are two words. Measured: well-known reports two words, well and known.
• Digits are words. 2026 is counted and listed like any other word.
• Quotation marks stay attached. 'hello' in single quotes is one token, listed with its quotes, and counted separately from the bare word.

So this is a usable frequency counter for plain English prose and a misleading one for anything else. If your text has accents or is not in a Latin script, the numbers on this page are wrong and nothing will tell you so.

The apostrophe your phone inserts splits the word in half

There are two apostrophes in common use. The straight one, U+0027, is what a plain keyboard produces. The curly one, U+2019, is what phones, word processors and most publishing software insert automatically. They look nearly identical and this tool treats them completely differently.

Measured, same word twice each time:

• don't with a straight apostrophe: “2 words, 1 unique”, listed as don't. Correct.
• don’t with a curly apostrophe: “4 words, 2 unique”, listed as don twice and t twice.

This is the single most likely reason a count here looks too high. Any text copied out of a word processor or a website will be full of curly apostrophes, and every contraction in it becomes two words, one of which is a stray t, s, re or ll near the top of your frequency list. If you see those in your results, that is what happened. Find-and-replace the curly apostrophes with straight ones first.

This tool and our Word Counter will give you different totals for the same text

They disagree by design, because they answer different questions, and we would rather say so than let you find out by accident. Measured on one sentence, It's a well-known café — naïve readers don’t notice 2026.:

• This tool: 12 words. Because café becomes caf, naïve splits into na and ve, don’t splits into don and t, and well-known splits into well and known.
• Word Counter: 10 words. Because it counts runs of non-whitespace, so each of those stays a single word, and the lone em dash counts as a word of its own.

Neither number is the “real” one. Use this tool when you want to know which words repeat, and Word Counter when you want a total to report. Do not quote the two together as though they measure the same thing.

The Top list is silently clamped between 1 and 100

Whatever you type in the Top box, the list you get is between one and a hundred entries, and nothing says so on screen. Measured against a text of 150 distinct words:

• Ask for 500 and you get 100.
• Ask for 0, or a negative number, and you get 1.
• Leave the box empty and you get 20, the default.

The summary line above the list is not clamped, so “150 words, 150 unique” keeps telling you the truth about the whole text even when the list below shows a hundred rows. If you need the full table for a long document, this is not the tool.

Words that tie are ordered alphabetically, not by where they appear

The list sorts by count, highest first, and breaks ties alphabetically using the browser’s locale-aware comparison. Measured: zebra apple mango, each appearing once, lists as apple, mango, zebra.

That is worth knowing because a long tail of words that all appear once will look like a meaningful ranking and is really just the alphabet. Only the top of the list, where the counts actually differ, carries information.

What this tool deliberately does not do

• No stop-word removal. the, a and of will dominate every result from ordinary prose. That is not a fault, it is what word frequency in English looks like, and you have to read past it.
• No stemming or lemmatising. run, runs and running are three separate words. So are singular and plural.
• No phrases. It counts single words only, so it cannot tell you that “climate change” appears forty times.
• No case distinction. Everything is lower-cased first, so Apple the company and apple the fruit are one entry.
• No export. There is no CSV or download; select the list and copy it.
• No reading of files. Paste your text in; the tool cannot open a document.

Nothing is uploaded. The counting is a few lines of JavaScript running in your browser, and your text never leaves your device.

Sources

• Why a literal a-z class excludes accented letters, and what it matches: regular expression character classes. developer.mozilla.org
• The Unicode-aware alternative this tool does not use, \p{L}: Unicode character class escapes. developer.mozilla.org
• Where word boundaries actually fall, in any language: Unicode Standard Annex #29. unicode.org/reports/tr29
• The two apostrophes, U+0027 and U+2019: Unicode General Punctuation chart. unicode.org/charts
• The locale-aware comparison used to break ties: localeCompare. developer.mozilla.org
• Lower-casing before counting, and its locale caveats: toLowerCase. developer.mozilla.org

Every behaviour and every number on this page was measured against this tool as it is published, in a browser, on 14 October 2026. Where the page says what the tool does not do, that is a statement about this tool, not about word frequency analysis in general.