Is my text uploaded or stored anywhere?

No. All four tools run in JavaScript in your browser. Your text is never transmitted, never logged and never stored, which is what makes them usable on an unpublished draft, a customer email list or an internal allowlist. Load a page, disconnect from the network, and the counts keep updating as you type.

Why do two tools give different word counts for the same text?

Because they disagree about what a word is. These counters split on whitespace, so 'well-known' is one word, '1,200' is one word and 'e.g.' is one word. Some editors treat an em dash written without spaces as a separator and count 'word—word' as two. Word and Google Docs also exclude some footnotes, headers and captions from their main count, and Word counts tracked deletions that are still displayed. The three counters here share a single word-splitting routine, so they always agree with each other. When a limit is being enforced, count in the tool the person checking will use.

The character counter shows two figures. Which one am I meant to use?

Almost always the first, the total with whitespace included — that is the number every platform cap is measured against, and a limit that excluded spaces would say so. The second figure is for text priced by volume rather than by hour: translation, transcription and typesetting are quoted per unit of text, and whether spaces are billed moves an invoice by roughly 15 to 18 per cent on ordinary English prose. One thing to watch is that it strips tabs and line breaks as well as spaces, so on a document of many short paragraphs the two totals diverge further than you would expect.

Why does the sentence counter split 'Dr. Chen arrived.' into two?

It splits on terminal punctuation, so any abbreviation ending in a period adds a spurious boundary — 'Dr.', 'e.g.', 'etc.', 'Inc.', 'U.S.'. URLs and three-part version numbers do the same. Ellipses and ordinary decimals are handled: runs of two or more periods are collapsed first and decimals are masked, so 'Wait... what?' counts as one and 'The rate rose to 4.5. That surprised nobody.' counts as two. On text dense with abbreviations, read the sentence count as an upper bound.

Can the list comparison find duplicates inside a single list?

No. The comparison is set-based: duplicates within a list collapse to one entry before anything is compared, so the tool tells you whether an item appears, never how many times. It is also exact-match only, so 'Acme Ltd' and 'Acme Ltd.' are two different items even with 'ignore case' switched on.

Text Tools

Four text tools that count as you type. There is no button to press, no sign-up and no upload — the counting happens in your browser, so a draft you have not published and a list of customer emails are both safe to paste in. Three of them measure the same text in different units, and picking the wrong unit is the usual reason a piece of writing sails past a limit it was checked against. The fourth answers a different question entirely.

Character Counter

Count the number of characters in your text.

Compare Lists

Compare two lists to find common and unique items.

Sentence Counter

Count the number of sentences in your text.

Word Counter

Count the number of words in your text.

Which Counter Do You Need?

Tool	What it measures	Usually checked against
Word counter	Words, with character and sentence totals alongside	Essay and article targets — a 650-word Common App essay, a 1,500–2,500-word academic essay, a 1,000–2,000-word post
Character counter	Characters, both including and excluding whitespace, plus words	Hard platform caps — 280 on X, 160 per SMS segment, a title tag near 60, a meta description near 155
Sentence counter	Sentences, with word and character totals alongside	Average sentence length — 15 to 20 words for a general audience
Compare lists	Membership across two lists: only in the first, in both, only in the second	Another list

The three counters overlap on screen — each shows more than one figure — but they are not interchangeable, because the three numbers behave completely differently when you are close to a limit.

A character cap is a hard ceiling. It is enforced by software, and going one over means rejection rather than disapproval. It is also the number most likely to surprise you, because the count is of UTF-16 code units rather than of things that look like characters. A plain emoji is 2. A hand with a skin-tone modifier is 4. The four-person family emoji is 11, because it is four emoji stitched together with three invisible joiners. SMS is stricter still: 160 characters fit in one segment only while the text stays inside the 7-bit GSM alphabet, and a single curly apostrophe pasted out of Word re-encodes the whole message and drops capacity to 70. The character counter is the tool to use whenever a machine, not a person, is doing the checking.

A word target is a guideline with one important exception. Ranges — 1,500 to 2,500 words for an essay, 1,000 to 2,000 for an article — are advice about depth, and landing at 1,480 costs you nothing. Stated maximums are different: the Common App simply stops accepting text at 650 words. Treat a range as guidance and a maximum as a ceiling, and use the word counter to check against either. What word count cannot do is tell you whether the writing is any good. There is no length that makes a page rank, and padding toward a number is visible to every experienced marker.

Average sentence length is a diagnostic, not a limit. Nothing rejects a draft for averaging 34 words a sentence; it just gets read badly. Divide the word count by the sentence count in the sentence counter and you have the single most actionable number in readability — it is a direct term in Flesch Reading Ease, in Flesch-Kincaid grade level and in Gunning Fog, and it is the only one of those inputs you can change with a full stop. Splitting a 300-word passage from 15 sentences into 20 lifts the Reading Ease score by about five points without touching a word of vocabulary.

The list comparison is doing something else altogether: it answers "what is in one of these and not the other?" — subscribers against buyers, a deployment allowlist against an infrastructure inventory, primary keys before and after a migration.

Counting Rules That Trip People Up

Every count is the output of a rule, and the rules are where surprises live.

Words are separated by whitespace, and nothing else. So "well-known" and "state-of-the-art" are one word each, 1,200 is one word, and 2026-07-26 is one word. This matches Word, Google Docs and most academic requirements. It also means the count is meaningless for Chinese, Japanese and Thai, which are written without spaces between words — count characters there instead.

Spaces count, and "excluding spaces" removes more than spaces. The second figure on the character counter strips every whitespace character, so tabs and the line breaks between paragraphs come out too. It is the right number for text priced by volume and the wrong number for any platform limit.

Sentence splitting is punctuation-based, with two fixes built in. Runs of two or more periods are collapsed before splitting, so an ellipsis does not become three sentences, and decimal numbers are masked, so "4.5" does not become a boundary. Stacked terminators like "?!" count once. Abbreviations, URLs and version numbers are not solved, and each one inflates the count by roughly one.

List comparison normalises before it compares. Trim whitespace and skip empty lines are on by default; the first also strips the stray carriage return that Windows line endings leave behind, which is the usual cause of two lists matching on nothing at all. Ignore case is off by default — turn it on for email addresses and hostnames, leave it off for case-sensitive IDs and API keys. Whitespace inside an entry is never touched, so Acme Ltd with two internal spaces will not match Acme Ltd.