You have two numbers for one document and you need to know which one your marker will see. Usually the answer is that the two tools were not measuring the same thing, and the difference disappears once you compare the right pair of figures. Occasionally it is something more interesting.
The causes below are ordered by how often they turn out to be the explanation, with a rough sense of how large a gap each one produces. That matters: if your two numbers are three hundred apart, there is no point investigating a cause that moves things by five.
By far the most common. Word counts the entire document; a checker with exclusions turned on counts the document minus the reference list, the citations, or whatever else you asked it to leave out. These are not competing answers to one question, they are answers to two. Compare Word's figure against the total rather than against the counted figure, and the gap usually resolves on the spot.
The number in Word's status bar is not the whole document. Footnote and endnote text is excluded from the default count, and so is anything in a header or footer. In a footnote-referencing discipline that is a substantial quantity of real writing that neither you nor Word is counting, which matters if your brief counts it.
Pasting into a plain-text box drops what a document carries but text does not: table cells may arrive as a run-on line or not at all, text boxes are usually lost entirely, footnote text does not travel, and captions become indistinguishable from body paragraphs. The count is not wrong so much as taken on a different document.
A .docx stores the word count Word calculated when it was last saved by Word. If the file has since been edited by Google Docs, LibreOffice, Pages or an export pipeline, that stored figure can be left behind — sometimes badly. It is a record of a past state, not a measurement of the current one.
What counts as one word is a decision, and different tools decide differently about hyphens, dashes, numbers, symbols and non-Latin scripts. This is real, and it is almost never the explanation for a gap of more than about one per cent. If your two numbers are 2,180 and 2,510, this is not why.
Most people assume a word count splits on spaces. Word’s rules are more particular than that, and the differences are the kind you would never guess. The examples below were measured against Microsoft Word itself rather than inferred, and StudyRef’s own counter is built to follow the same rules — each one is pinned by a test, so the counter cannot drift away from them quietly.
| Text | Counts as | Why |
|---|---|---|
| cost-benefit analysis | 2 words | A plain hyphen does not separate words. The hyphenated compound is one word, which is why a heavily hyphenated document counts lower than you might expect. |
| cost–benefit analysis | 3 words | An en dash does separate them, even with no spaces around it. The two examples look nearly identical on screen and differ by a word. |
| revenue 2020-2021 total | 3 words | Same rule with numbers: a hyphenated range is a single token. |
| revenue 2020–2021 total | 4 words | And the en-dash version is two. A style guide that asks for en dashes in date ranges will therefore raise your word count slightly, at no cost to you. |
| Smith & Jones | 3 words | A standalone ampersand counts as a word, as does a standalone hyphen. Isolated punctuation is not automatically ignored. |
| • word | 1 word | The bullet glyph is the exception. Word does not count it even when you typed it as ordinary text rather than using a list. |
| https://example.com/a/b | 1 word | A URL or email address is one word. Splitting on punctuation would make a bibliography full of links score far higher than it does. |
| 안녕하세요 저는 학생입니다 | 12 words | Korean, Chinese and Japanese are counted per character rather than per space-separated run — three visible words here, twelve counted. The same document in English and in Korean will not produce comparable numbers. |
None of these will move a 2,000-word essay by more than a few words. They are worth knowing because they explain why a small residual difference between any two counters is normal and not a defect in either of them.
When you upload a .docx, there are two candidate totals: StudyRef’s own fresh count of the document’s contents, and the figure Microsoft Word saved inside the file the last time it wrote it. StudyRef normally prefers its own, because it is a measurement of the document as it is now rather than a record of what it once was.
It defers to Word’s saved figure in one situation: when the file contains a structure StudyRef’s counter has not been independently verified against — equations, for example. In that case Word’s number is the more trustworthy of the two, and the honest thing is to use it and say so. Either way the result tells you which of the two you are looking at, so a difference is explained rather than left as a mystery.
What StudyRef does not claim is an exact match with Word on every document. No tool can promise that across arbitrary files, and a product that did would be asking you to trust a claim it has no way to keep. Preferring Word’s own figure where StudyRef’s is unproven is what that honesty looks like in the implementation rather than in the marketing.
Put Word's figure next to the other tool's total, not next to a figure that already has exclusions applied. If they now agree, there was never a discrepancy — only two different questions.
If your document uses footnotes for referencing or commentary, Word's status bar has been quietly leaving them out the whole time. Select the body text and check the selection count against the whole-document count to see the size of it.
If one of your two numbers came from pasting, redo it on the .docx. Tables, text boxes and footnotes are the usual casualties, and all three are invisible losses — nothing warns you that the paste was incomplete.
If the document has been through Google Docs or an export, treat the count stored inside it as unreliable and prefer a fresh measurement of the actual content.
Below about one per cent, the remaining gap is tokenising, and no amount of investigation will close it because there is no single correct answer. It is not the difference that decides a penalty.
See the total, the exclusions and the remainder together
Upload the .docx and StudyRef shows all three numbers, which category removed what, and whether the total is its own count or the one Word saved with the file. Free, and no account.
Count my document