Why word counts differ
Two counters can disagree because they count different parts of a document or divide the same text into words differently. Check the text being counted first, then check the word boundaries.
Google Docs, Word and Taskono: what can you compare?
A document editor knows about document regions. A plain-text counter sees the text you give it. Pasting a footnote into a text box makes it ordinary text; its original document role is lost.
| Counter | What is documented or implemented | What to check |
|---|---|---|
| Google Docs | Its help page says the whole-document count excludes headers, footers and footnotes unless you select a section. Open Tools → Word count to see the count.Google Docs Help | Check the selected text and document regions. The cited help page does not specify exact hyphen, emoji or multilingual token counts. |
| Microsoft Word | Microsoft documents counts while typing and a selection count alongside the document count in the status bar. Its help page gives separate Windows, macOS and web instructions.Microsoft Support | Record your Word version, platform, selected text and any count-dialog options. The cited help page does not give an exact tokenization rule for these examples. |
| Taskono Word Counter | Counts the entire editor text, including any pasted title, citation or note. Selecting part of the editor does not restrict the count. Text analysis runs in your browser.Try Taskono Word Counter | Paste only the required scope. Standard segmentation and the older-browser fallback can differ; the examples below show both. |
Try these original test examples
These are results from Taskono’s actual counting function, not Google Docs or Word measurements. The standard column uses Intl.Segmenter; the fallback column forces the alternative path used when that API is unavailable. Both columns are checked against fixed expected results in tests. Standard results are also checked in the live editor across the project’s Chromium, Firefox and WebKit browsers.
In the whitespace sample, the gap after “one” is a tab followed by a space; “three” begins on a new line. The hidden-space sample contains U+200B between alpha and beta. You can select and copy each sample exactly as displayed.
| Exact sample and explanation | Standard | Fallback |
|---|---|---|
Re-check the well‑made plan.Hyphen-minus and non-breaking hyphen both split the compounds. | 6 | 6 |
We can't; they don’t.Straight and curly apostrophes remain inside these contractions. | 4 | 4 |
Item 42 costs 3.50.42 counts as a word. The standard path keeps 3.50 together; the fallback splits it. | 4 | 5 |
Wait... really?!Punctuation marks do not add words. | 2 | 2 |
one two
threeExtra whitespace does not create extra words. | 3 | 3 |
Hi 👩💻!Only Hi contributes a word. The joined emoji contributes characters, not words. | 1 | 1 |
Привет мирTwo Cyrillic words separated by a space. | 2 | 2 |
你好世界Standard segmentation finds two words in this Chinese sample; the fallback keeps the string together. | 2 | 1 |
café caféThe first accent is precomposed; the second uses a combining mark. Each spelling contributes one word. | 2 | 2 |
alphabetaAn invisible zero-width space separates the two words. | 2 | 2 |
What counts as a word in Taskono?
Taskono normalizes Windows and older Mac line endings to newline characters. It segments sentences within each physical line, then counts word-like segments with the browser’s Intl.Segmenter using the English locale. It does not count by simply splitting on spaces.
Before word segmentation, it replaces the ordinary hyphen-minus (-), hyphen (‐, U+2010) and non-breaking hyphen (‑, U+2011) with spaces. Those characters therefore split compounds. Straight and curly apostrophes can stay inside contractions. Numbers can be word-like segments, and a decimal such as 3.50 stays together in the standard example below.
Punctuation marks and emoji alone do not add words. Spaces, tabs and newlines separate text without adding empty words. A hidden zero-width space can also create a boundary even though it looks like no space at all.
Cyrillic, Arabic and other scripts containing letters are eligible for counting. For scripts without spaces, such as Chinese, Intl.Segmenter can find boundaries inside an unbroken string. This is a browser segmentation rule, not a guarantee of agreement with a teacher, editor or every language’s linguistic definition of a word.
If Intl.Segmenter is unavailable, Taskono uses a Unicode letter-and-number pattern, permitting combining marks and internal straight or curly apostrophes. Its simpler sentence split treats periods as boundaries, so the decimal example below gains a word. A continuous Chinese string is one word under this fallback. Browser and Unicode-data versions can affect segmentation.
A character count answers a different question
Taskono counts Unicode code points after normalizing line endings. A joined emoji can look like one symbol while containing several code points. An accented letter can be stored as one precomposed character or a letter plus a combining mark. Neither changes the word count in the examples above.
The “characters without whitespace” result removes Unicode whitespace and the byte-order-mark character. Zero-width spaces and joiners are not removed by that rule. Comparing words in one app with characters in another will not diagnose a word-count discrepancy.
How to resolve a difference
Start with the required submission rules: which application or document regions does the recipient expect? If a teacher or editor specifies a counter, use that requirement for the final count.
Make a copy of the same plain text for each counter. Check titles, citations, footnotes and selections before changing punctuation. In Taskono, paste only the passage you want counted into an empty editor.
If totals still differ, isolate one sentence at a time. Try a hyphenated compound, a contraction and a decimal number separately, then check hidden whitespace and text without spaces. Record the exact sample, app version, browser and result so the comparison can be repeated.
Do not silently rewrite a document to make two counters agree. A documented difference is more useful than a matching total produced by changing the text.
Put it into practice
Paste a sample into Word Counter to see its words, characters and other statistics. The tool page explains the controls, goals and optional browser saving.
Sources and verification
The official help pages above establish how to find counts and select their scope. They do not establish exact cross-product results for our sample strings. Taskono’s examples were verified on November 2, 2026; results are tied to the implemented rules and tested browser versions.