The phrase "Hello, world." contains thirteen characters, twelve without spaces, ten letters, two words and one sentence. None of those is the right answer on its own — which count you want depends entirely on who is asking, and the gap between them is wide enough to fail a limit you thought you were inside.
Most social platforms and SMS count characters including spaces. Academic word limits count words. Typesetting cares about characters without spaces. Picking the wrong one is not a rounding error.
How do the counts differ?
On the same short phrase, the spread is wide.
| Measure | "Hello, world." |
|---|---|
| Characters with spaces | 13 |
| Characters without spaces | 12 |
| Letters only | 10 |
| Words | 2 |
| Sentences | 1 |
Letters is the strictest count — spaces, digits, punctuation and symbols all excluded — which makes it consistently the smallest number and the one least often asked for outside word games and puzzles.
Why is sentence detection approximate?
Because it works from terminating punctuation, and full stops do other jobs. Abbreviations end in one, decimals contain one, and an ellipsis is three — so "Dr. Smith arrived at 3.30 p.m." reads as several sentences to any punctuation-based counter.
It is the standard approach because the alternative requires understanding the text. For ordinary prose the error is small; for anything dense with abbreviations or figures it is not, and the count should be read as an estimate.
The useful output is not the sentence count but the ratio. Words divided by sentences gives average sentence length, and that number is a far better guide to whether a paragraph is readable than any individual count.
Which line convention is being used?
Two collide and it matters which. An editor counts every line break as a new line, blank lines included. A paragraph counter treats a block separated by a blank line as one paragraph and ignores the blanks.
Between them you can tell how many paragraphs a document really has and how much of its length is spacing. Text using a single newline between paragraphs rather than a blank line confuses the second convention, which is why pasted content sometimes reports one enormous paragraph.
Comparing paragraph count against sentence count is the quickest structural read available: twenty sentences in three paragraphs reads very differently from the same twenty in twelve.
Can syllables be counted reliably?
Not exactly, and no algorithm manages it. The standard approach counts vowel groups and discards a silent trailing "e", which is right for the large majority of ordinary English words and wrong for a predictable set.
English spelling does not determine syllable count — "fire" is one syllable or two depending on the speaker, "every" is commonly two despite looking like three, and any word ending in "-le" after a consonant breaks the rule. The estimate is useful in aggregate and unreliable on any individual word.
That is enough for readability formulas, which average over hundreds of words, and not enough for poetry, which is exactly where people want it.
What counts as a word?
A run of letters or digits separated by whitespace, in most implementations — so hyphenated and apostrophised words count once. "Well-known" is one word, "don’t" is one, and "state of the art" is four.
That definition is why word counts differ between tools on the same document. Some treat a hyphenated compound as two, some count numbers and some do not, and text in footnotes, headers and captions is included by some and excluded by others. For an academic limit the only definition that matters is the one in the submission guidance.
What is frequency analysis for?
Ranking every distinct character by how often it appears — the oldest technique in cryptanalysis, and still enough to break a simple substitution cipher by inspection. In English text E accounts for around twelve per cent of letters, and the ordering of the top few is stable.
The practical modern use is spotting what a document is made of. An unexpected character near the top of the list usually means an encoding problem, a smart quote where a straight one was intended, or a non-breaking space that will not wrap.
Spaces and line breaks are worth including and labelling rather than dropping, because they are frequently the answer to why a count is higher than expected.
Questions people ask
Is Y a vowel? No rule settles it. Counting it as a consonant is the schoolroom convention and it is wrong for "rhythm" and "myth". Any tool has to pick one and say which.
Do accented letters count as their base letter? Usually yes — é counts as a vowel and as a letter. Whether it is one character or two depends on the encoding, which the UTF-8 article covers.
Why does my word count differ from Word’s? Different definitions of a word. Hyphenated compounds, numbers, and text in footnotes or headers are handled differently by different tools, and none of them is wrong.
Does any of this leave my browser? No. Counting is arithmetic on the characters and happens on the page.
Five counts, five questions, and it is worth knowing which was asked. The character counter shows both character figures at once, the letter, vowel, sentence, line and paragraph counters each isolate one, the syllable counter gives the estimate readability formulas run on, and the character frequency counter ranks what the text is actually made of.