Counting words sounds like it should have one right answer. It does not, and the reason matters if you are writing to a limit: two reputable tools can look at the same paragraph and disagree by several per cent, entirely legitimately, because they resolve the ambiguous cases differently.
The convention this tool uses
Text is split on whitespace — spaces, tabs, line breaks — and each resulting non-empty token counts as one word. That is the most widely used definition and the one most word processors approximate. It produces these specific outcomes:
- Hyphenated compounds count as one. “State-of-the-art” is one word, because there is no whitespace in it.
- Contractions count as one. “Don’t” and “we’ll” are single words.
- Numbers count as words. “1,500” is one word; “1 500” is two.
- Standalone symbols count. An isolated “&” or “—” surrounded by spaces counts as a word.
- Multiple spaces collapse. Double spaces, tabs and line breaks do not inflate the count.
- Em dashes inside words do not split them. “Well—perhaps” counts as one, which is a case where reasonable tools genuinely differ.
Why your word processor disagrees
If your count here differs slightly from elsewhere, one of these is usually why:
Footnotes, headers, captions and endnotes. Word processors may include or exclude these depending on settings. Pasting the body text alone gives a different number from the document total.
Fields and auto-generated content. Page numbers, tables of contents and citation fields may or may not be counted.
Em-dash and slash handling. Some tools treat “and/or” as two words; splitting purely on whitespace treats it as one.
Tracked changes. A document with pending edits can count either version.
The practical rule: if a limit is being enforced, use the counter the enforcer uses. For a university assignment that is the university’s stated method; for a publisher it is whatever their submission guidelines say. Any other counter is an estimate, and being 40 words over because of a definitional difference is a bad way to find this out.
Rough equivalents
Useful when you are planning rather than checking:
| Length | Words | Reading time |
|---|---|---|
| Tweet / short post | 50–70 | 15 sec |
| 100–200 | under 1 min | |
| Blog post | 600–1,200 | 3–5 min |
| Long-form article | 1,500–3,000 | 6–12 min |
| A4 page, single-spaced | ~500 | 2 min |
| A4 page, double-spaced | ~250 | 1 min |
Reading time here assumes about 240 words per minute, which is a reasonable average for adult silent reading of general prose. Technical material runs slower; skimming runs much faster.
Where people go wrong
Writing to the limit and pasting later. Formatting, citations and headings can shift the count. Check in the environment that will be assessed.
Assuming a character limit and a word limit are interchangeable. English averages roughly 5 characters per word plus a space, so 500 words is around 3,000 characters — but “roughly” is doing real work there, and technical vocabulary breaks the estimate badly. If the limit is in characters, use the character counter.
Counting a draft with placeholder text still in it. Obvious, and it happens constantly.
Treating the count as a measure of quality. Hitting 1,000 words says nothing about whether the piece is any good, and padding to reach a limit is visible to every reader. The readability score is a better signal, and still not a verdict.
Your text is processed entirely in your browser and never transmitted — see the privacy policy.