Why Character and Word Counters Give Different Results

Differences are normal

Paste the same text into three counters and you may get three slightly different results. This is rarely a bug. Counting text involves decisions about edge cases, and each tool makes its own. Once you know the common decisions, you can usually explain a difference in a minute.

The size of the gap is a clue. A difference of one or two usually points to a line break or a trailing space. A difference of ten or more usually points to emoji, hyphens or a different definition of a word.

Spaces and line breaks

Some tools count a line break as one character and some as two, because on older systems a break was stored as a carriage return plus a line feed. Some tools trim the spaces at the start and end of the text, and others count them. A text copied from a document often ends with a hidden line break.

To compare tools, test with a very short text and see what each tool reports for “a b” and for the same text with a line break between the letters.

Emoji and special characters

A tool built on UTF-16 length counts an emoji as two characters, and a family emoji as eight. A tool that counts what you see reports one. Platforms such as X apply weights on top of that. If a text contains emoji, expect differences.

Hidden characters make a difference too. A zero-width space or a non-breaking space looks like nothing or like an ordinary space, and some tools count it and some ignore it.

What counts as a word

The usual rule is that a word is a group of characters between spaces. Tools diverge on hyphens, slashes, dashes with spaces around them and lone symbols. “Well-known” may be one word or two. “And/or” may be one or two. A dash standing alone may be counted or not.

Numbers and dates are another source. Most tools count them, and a few skip pure numbers. When the target is a hard limit, the safest approach is to count the way the recipient does.

Languages without spaces

For Japanese, Chinese and Thai there are no spaces to split on, so the word count depends on a segmentation method. A tool that splits on spaces reports the whole sentence as one word. A tool that uses a dictionary-based method reports a plausible count. Different dictionaries give different answers.

For these languages the number of characters is the more stable measure, and many editors and publishers use it for that reason.

Sentences and paragraphs

Sentence counts differ most. A tool that splits on full stops counts “Mr. Smith left.” as two sentences, and a tool with a list of abbreviations counts one. Decimal numbers, web addresses and ellipses cause similar problems. Paragraph counts differ in whether a single line break or a blank line starts a new paragraph.

These figures are best seen as estimates. They are useful for comparing drafts but not exact enough for rules.

How to decide which number to trust

First, find out who sets the limit. If a platform or school has its own counter, use it as the authority. If you submit a file, the word processor of the recipient is the likely reference. Second, count the way the rule says: with or without spaces, with or without footnotes.

If there is no authority, use a tool that explains its method, and keep a margin of a few per cent below any hard limit. Counting with the same tool for the whole project keeps your figures consistent.

A test you can run in a minute

To see how a tool counts, paste five short samples and write down the results: “a b”, “a b” with two spaces, “well-known”, “café 👍” and a text of two lines. Compare the figures with the explanations above. Within a few minutes you know how the tool treats spaces, hyphens, accents, emoji and line breaks.

Run the same samples through the tool that sets your limit. If the results match, you can trust the figures. If they differ, you know exactly which rule to correct for.

When the difference matters

For most writing, a difference of a few characters is irrelevant. It matters when a limit is hard: a form that refuses longer input, an advertisement platform, a printed label, a database field or a competition entry with a strict maximum.

In those cases, count in the tool that will judge your text, leave a margin, and avoid writing right up to the limit. A text that is five per cent under the limit passes in every tool.

How this counter decides

This counter counts visible characters, so an emoji is one. Line breaks count as one character each. A word is a group of letters or digits between spaces, and Japanese and Chinese words are found with the browser’s segmentation. A sentence ends at a full stop, a question mark or an exclamation mark followed by a space or the end of a line, or at their Japanese forms.

A paragraph is a non-empty line, and the line count includes blank lines. The description in the footer of the tool states the main rules, so you can compare it with other tools and see why a figure differs.