Character and Word Counts: Why Counters Disagree

Last updated: 2026-10-11 · Checked and published by Leo, who runs Hanji

Spaces, line breaks, bytes, emoji and reading time: how character and word counts work for essays, abstracts and posts, and why two counters differ.

You paste the same essay into two counters and get two different numbers. Usually neither is wrong; they are answering slightly different questions. Is a space a character? Is a line break? Is “well-known” one word or two? Is 👍🏽 one character or four? This guide explains the choices behind each number so you can match the one your form, teacher or platform uses.

Characters with spaces, without spaces, and line breaks

Take this two-line paragraph, which we ran through Hanji’s word counter:

I learned to check a problem with numbers before trying to fix it. During my internship I sorted 1,200 support tickets, found the three most common complaints, and rewrote the help text so those tickets fell by 30% in a month.

Count Result What it includes
Words 41 Groups of characters separated by spaces or line breaks
Characters 225 Letters, digits, punctuation and spaces; not the line break
Without spaces 186 Neither spaces nor line breaks
With line breaks 226 Everything, the line break counted as one

In English, spaces make up a large share of the text, so “with spaces” and “without spaces” differ a lot: here by 39, about a sixth. If an application asks for “500 characters” and does not say which, assume spaces are included: the length limit built into web form fields (the maxlength attribute) counts every character, spaces included. If it matters, ask.

Line breaks are the quiet troublemaker. Our “Characters” figure leaves them out. A web form, though, may count each one, and there is a technical twist: when a browser submits a form, the HTML standard has it send every line break as two characters (a carriage return and a line feed), even though the field’s own length limit counts each one as one. So a counter on the page can say you are exactly at the limit while the server that receives the text sees a few characters more, one extra for every paragraph break. If a form rejects text that “should” fit, remove blank lines between paragraphs or leave a margin of a few characters.

How words are counted

Nearly every tool counts a word as a run of characters between spaces. That is simple and predictable, and it is what we do. It also means:

  • Hyphenated words count once. “Well-known” and “state-of-the-art” are one word each.
  • Numbers count as words. “1,200” and “30%” each count once.
  • Dashes depend on spacing. “the result—a 30% drop” has a dash with no spaces, so “result—a” is one word. Write “result — a” and the dash itself becomes a word.
  • Links and email addresses count once, however long.
  • Text pasted from a PDF can carry broken-up words (“informa- tion”) or odd spaces from the original layout, which change the count.

Other programs may make different calls on hyphens, dashes, numbers and symbols, so word processors and online counters can differ by a few words on a long essay. For a limit, the count that matters is the one the person or system checking it uses: if a teacher sets “1,500 words”, ask which program they check with; a submission portal goes by its own counter.

Words are also a poor measure for languages that do not put spaces between words. Japanese text with no spaces counts as one “word” per line, so for Japanese, and for Korean where a space-separated unit is often several words long, go by the character counts.

What counts as one character

We count what you see as a character. An emoji such as 👍🏽 (thumbs up with a skin tone) or a flag such as 🇰🇷 is stored as two or more code points but shows as one symbol, so it counts as one.

Not every system agrees. Web forms and scripts often count the way JavaScript measures string length, in UTF-16 code units; a form field’s built-in maxlength limit counts that way too. For the short text “Great 👍🏽 🇰🇷” we count 9 characters; a form counting UTF-16 units would see 15, because each of those two emoji takes four units. Accented letters can also differ: “é” can be stored as one code point or as “e” plus a combining accent, and a naive counter sees two.

If your text has emoji and a form says it is too long when your counter says it fits, this is almost always why.

Bytes: when a limit is about storage, not reading

Some systems limit text by bytes because that is how a database column is sized. In UTF-8, the encoding used by about 99% of websites (W3Techs, October 2026):

  • English letters, digits, spaces and basic punctuation: 1 byte each
  • Most accented Latin letters, and Greek, Cyrillic, Arabic or Hebrew letters: 2 bytes
  • Korean, Japanese and Chinese characters: 3 bytes (a few rare Chinese characters take 4)
  • Most emoji: 4 bytes, and a skin-tone or flag emoji 8 or more

For our English paragraph the UTF-8 size is 226 bytes, the same as the character count with the line break, because every character is a 1-byte one. For a Korean paragraph of 111 characters it is 255 bytes. Older Korean and Japanese systems instead count every non-English character as 2 bytes; the counter shows that figure too.

Limits for essays, abstracts and posts

  • Essays and personal statements. Read the instructions for what is included: the title, footnotes, references, quotations and headings are counted by some and excluded by others. If the limit is a range (“about 1,500 words”), check whether going over is penalised.
  • Abstracts. Journals and conferences set their own limits. The author guidelines give the number and say whether the title and keywords count.
  • X (Twitter). A post can be up to 280 characters on a standard account, but X weights them: most Latin letters count as 1, Korean, Japanese and Chinese characters as 2, an emoji as 2, and every link as 23 however long it is. So a standard post in Japanese holds about 140 characters. Our counter applies these weights. X Premium subscribers can post longer texts.
  • Meta descriptions. Google sets no length limit for the text under a search result. It shortens snippets as needed to fit the device width, and it often builds the snippet from the page content instead. Put the important part first.

For other platforms, check the limit in the app itself; they change, and they often count emoji and links in their own way.

Reading and speaking time

Reading time in our counter assumes about 238 words a minute, the average a 2019 review of 190 studies found for adults silently reading English non-fiction. Speaking time assumes 130 words a minute, a deliberately unhurried pace. For a 1,000-word essay that gives 4 min 12 s to read and 7 min 42 s to say out loud. Both are averages: technical text reads slower, and nervous speakers go faster than they think. When you prepare a timed talk, rehearse it once with a timer and use the estimate only to plan the first draft.

Manuscript paper pages (Korean and Japanese)

On the Korean and Japanese versions of the counter you will also see manuscript paper pages: how many sheets of 20-square-wide paper (400 or 200 squares a sheet) the text would fill if written out by hand. Some schools and contests still set lengths this way; Japan’s national reading-essay contest for students, for example, sets its limits in 400-character sheets. Every paragraph starts on a new line, so the same 400 characters fill exactly one 400-square sheet as a single paragraph, but one sheet and four lines when split into eight paragraphs of 50. The Korean and Japanese versions of this guide explain the rules in more detail.

Why two counters disagree: a checklist

When the numbers do not match, check these in order:

  1. Spaces included or not
  2. Line breaks counted as 0, 1 or 2 characters
  3. Emoji and accented letters counted as one or several
  4. Hyphenated words, dashes and numbers in the word count
  5. Characters versus bytes, and which byte encoding
  6. Hidden extras from pasting: trailing spaces, tabs, non-breaking spaces, blank lines

The counting happens on your device and the text is never sent anywhere, so you can paste an unsubmitted application; see what happens to your files for how Hanji’s tools handle data.

Guides