How Word Counters Count: Words, Characters and Reading Time

A word counter treats a word as any run of characters separated by spaces or line breaks, which is why "well-known", "3.5" and "and/or" each count as one. Characters are counted individually, usually with and without spaces. Sentences are split at full stops, question marks and exclamation marks, and paragraphs at blank lines. Tools disagree because each makes slightly different decisions about hyphens, dashes, numbers, emoji and punctuation that stands alone.

What counts as a word

There is no universal definition. The one that nearly every piece of software uses is mechanical: split the text on whitespace, throw away any empty pieces, and count what is left. Under this rule the following are all single words:

• well-known (hyphen, no spaces)
• don't (apostrophe)
• 3.5 and 1,000,000 (numbers)
• and/or (slash)
• £250 and 50% (symbol attached to a number)
• email@example.com

And these count as two words each, because there is a space in the middle:

• ice cream
• well - known (hyphen with spaces around it: the hyphen itself is a third "word" in many tools)
• 3 . 5

Microsoft Word, Google Docs, LibreOffice and most online counters agree on all of the above. Where they part company is on characters that sit alone. An em dash with spaces on both sides ( — ) is counted as a word by Google Docs and by Word, because it is a run of non-space characters. A tool that only counts runs containing at least one letter or digit ignores it. The same goes for a lone ellipsis, a bullet character pasted from a web page, or a stray asterisk.

Why Word, Google Docs and online counters disagree

The differences are small, usually under two per cent, but they are real and they follow predictable patterns.

Case	Microsoft Word	Google Docs	Typical online counter
Hyphenated compound ("self-service")	1 word	1 word	1 word (some count 2)
Em dash with spaces ( — )	1 word	1 word	0 words (usually)
Number ("2024")	1 word	1 word	1 word (a few count 0)
Contraction ("it's")	1 word	1 word	1 word
Emoji on its own	1 word	1 word	0 or 1
URL	1 word	1 word	1 word (some split at slashes)
Footnotes and text boxes	Optional	Counted	Not applicable
Headers and footers	Excluded	Excluded	Not applicable

The last two rows explain the biggest surprises. Word's count excludes headers and footers and, by default, footnotes and endnotes, although there is a checkbox to include them. Google Docs counts footnotes. If an essay has a reference list in a footnote, the two programs can differ by hundreds of words on the same file. Hidden text, comments and tracked deletions are excluded in both, but accepted insertions are included, so a document mid-review can show a different number from the finished version.

Online counters work on plain text, so they never see footnotes or headers. What they see is exactly what you paste. If you copy from a web page, you often bring along hidden navigation text, image captions or a "Share this" line, and the count rises accordingly. Paste into a plain text field first and check that the text is what you intended.

The Word Counter on this site uses the whitespace rule, counts numbers as words, ignores runs that contain no letters or digits, and shows the character count alongside so you can see both measures at once.

Characters: with spaces, without, and the emoji problem

Character counts look simpler than word counts and are actually messier, because "character" can mean three different things.

• Characters with spaces counts every visible letter, digit, punctuation mark and space. Line breaks are usually counted too, each as one character. This is what most platform limits refer to.
• Characters without spaces removes spaces (and often tabs and line breaks) from the total. Translation agencies and some academic journals use this measure because it reflects the amount of typing rather than the layout.
• Bytes or code units are what the computer actually stores. An "é" is one character to a human and two bytes in UTF-8. A flag emoji is one picture and two code points; a family emoji can be seven code points joined with invisible joiners.

Platforms pick different measures. X (formerly Twitter) counts most Latin, Cyrillic and Greek characters as one, but counts emoji and most Chinese, Japanese and Korean characters as two towards its 280-character limit. An SMS message in the basic GSM alphabet holds 160 characters, but one emoji or curly quote forces the whole message into a different encoding that holds only 70. Instagram and Facebook count each emoji as one. There is no single right answer, which is why a good counter reports characters both with and without spaces and lets you check against the specific platform.

Use the Character Counter when a limit is specified in characters; it shows both totals and marks the common social and SEO limits.

Sentences and paragraphs

Sentence counting is the least reliable of the four measures because it depends on punctuation, and punctuation is ambiguous. The simple rule is: a sentence ends at a full stop, question mark or exclamation mark followed by a space or a line break. That rule breaks in predictable places.

• Abbreviations. "Dr. Smith arrived at 3 p.m. on Monday." has three full stops and one sentence. A naive counter reports three. Better counters keep a list of common abbreviations and ignore a full stop that follows one, but no list is complete.
• Decimals and version numbers. "3.5" and "v2.0.1" contain full stops that are not sentence ends. Most counters handle these by requiring a space after the stop.
• Ellipses. "I wonder... perhaps not." is one sentence to most readers and two to many counters.
• Headings and list items. They usually have no end punctuation, so they are either not counted at all or merged into the next sentence.
• Quotes. "He said 'Go.' and left." is one sentence containing a full stop inside quotation marks.

In practice a sentence count is accurate to within a few per cent on ordinary prose and should be treated as an estimate. It is still useful: average sentence length (words divided by sentences) is one of the quickest readability signals there is. Around 15 to 20 words per sentence reads comfortably; over 25 and most readers start to lose the thread. If you want more than a single average, the Readability Score Checker reports Flesch reading ease and several grade-level scores, names your longest sentence and counts how many run past 25 words.

Paragraph counting is more consistent. Almost every tool treats one or more blank lines as a paragraph break. The disagreement is over single line breaks: a word processor treats every press of Enter as a new paragraph, while a plain-text counter usually needs an empty line between paragraphs. Text copied from an email or PDF often has a line break at the end of every visual line, which inflates the paragraph count enormously until the breaks are removed.

The Sentence Counter reports sentences, words, characters and paragraphs together and calculates the average sentence length for you.

Reading time and speaking time

Reading time is not measured; it is calculated. The formula is simply:

reading time in minutes = word count ÷ reading speed in words per minute

The reading speed is the whole debate. The most cited modern figure, from a 2019 meta-analysis of nearly 200 studies, is an average of about 238 words per minute for adults reading non-fiction silently in English, and around 260 for fiction. Many blogs and publishing platforms round this to 200 or 250, and a few use 275 or 300 to make articles look shorter. Technical or dense material is read more slowly, nearer 150 to 200. Skimming on a phone can be much faster.

Word count	At 200 wpm	At 238 wpm	At 250 wpm	Spoken at 150 wpm
500	2 min 30 s	2 min 6 s	2 min 0 s	3 min 20 s
1,000	5 min 0 s	4 min 12 s	4 min 0 s	6 min 40 s
1,800	9 min 0 s	7 min 34 s	7 min 12 s	12 min 0 s
3,000	15 min 0 s	12 min 36 s	12 min 0 s	20 min 0 s

The arithmetic for one row: 1,800 words divided by 238 words per minute is 7.56 minutes. The fractional part, 0.56 of a minute, is 0.56 times 60, or about 34 seconds, giving 7 minutes 34 seconds. Most sites round to the nearest whole minute.

Speaking time uses the same formula with a slower rate. Conversational English runs at roughly 150 words per minute; a careful presenter or newsreader is closer to 130; an auctioneer or excited podcaster can exceed 200. For a speech or a voiceover script, 130 to 150 is the safe planning range. A five-minute talk therefore needs about 650 to 750 words, not the 1,000 or more that first-time speakers tend to write.

Images, tables and code add time that the formula ignores. A common adjustment, used by Medium among others, is to add about 12 seconds for the first image and a little less for each subsequent one. For most articles the difference is small enough to ignore.

Essay and application limits

Limits are where counting rules stop being academic. A few that writers meet regularly, correct at the time of writing:

• Common App personal essay (US universities): 250 to 650 words, enforced by the application form itself.
• UCAS personal statement (UK): 4,000 characters including spaces, now split across three structured questions rather than one free essay.
• IELTS Writing: at least 150 words for Task 1 and at least 250 for Task 2. There is no upper limit, but under-length answers are penalised.
• University essays and dissertations: typically a stated maximum with a tolerance of plus or minus 10 per cent, and most departments exclude the reference list, appendices and sometimes footnotes. Check your handbook, because some count everything.
• Journal abstracts: commonly 150, 250 or 300 words.
• Cover letters and job applications: often a character limit in an online form, frequently 2,000 to 4,000 characters.

Three practical rules follow from everything above.

1. Count in the tool the recipient will use. If an online form enforces a limit, the form's own counter is the only one that matters. If a university says "as counted by Microsoft Word", use Word, with footnotes included or excluded as the handbook says.
2. Leave a margin. Two per cent is enough to absorb the differences between counters. For a 2,000-word limit, aim for 1,950.
3. Know what is excluded. Titles, reference lists and figure captions are often excluded from academic limits and included by every online counter. Paste only the body if that is what will be measured.

Where to go next

Paste your draft into the Word Counter for words, characters and reading time at a glance, use the Character Counter when a form or platform specifies a character limit, and try the Sentence Counter to check average sentence length before you submit. Each tool runs in your browser and nothing you paste leaves your device.

Frequently asked questions

Do hyphenated words count as one word or two?

In Microsoft Word, Google Docs and most online counters, a hyphenated word with no spaces around the hyphen counts as one word, so "well-known" and "self-service" are each one word. If there are spaces around the hyphen, the parts are counted separately and the hyphen itself may be counted as a word. A few counters split on hyphens deliberately, which is why totals can differ by a handful of words between tools.

Do numbers count as words in a word count?

Yes, in nearly all word processors and counters. Any run of characters between spaces is a word, so "2024", "3.5" and "£1,000" each count as one. A small number of counters ignore runs that contain no letters, so numbers written as digits would not be counted. If a limit is strict and your text is full of figures, check it in the tool the recipient will use rather than relying on one counter.

Why does Google Docs give a different word count from Word?

The two agree on ordinary prose but differ at the edges. Google Docs includes footnotes in its count; Word excludes footnotes and endnotes unless you tick the option to include them. Both exclude headers and footers. They also handle dashes and stray punctuation slightly differently. On a long academic document with many footnotes, the gap can run to hundreds of words, so always use the counter your institution specifies.

How is reading time calculated?

Reading time is the word count divided by an assumed reading speed. Most sites use 200 to 250 words per minute; research on adult silent reading puts the average for English non-fiction at about 238. So a 1,200-word article is shown as five or six minutes depending on the speed chosen. Speaking time uses a slower rate, usually 130 to 150 words per minute, so the same article would take about eight to nine minutes to read aloud.

Does an emoji count as one character or two?

It depends on the platform. In a word processor or a standard character counter, one emoji is one character. X counts most emoji as two characters towards its 280-character limit. In an SMS, a single emoji switches the message to 16-bit encoding, which cuts the capacity from 160 characters to 70. Some emoji are built from several code points joined together, so the number a program reports can be higher still.

Does the word count include the title and references?

An online counter includes everything you paste, so it will count a title and a reference list if they are in the text. Academic limits usually exclude the title page, reference list, bibliography and appendices, and sometimes footnotes and tables as well, but policies vary between institutions and even between departments. Read the handbook for the specific assignment and paste only the parts that will actually be measured.