Checking a piece of writing, step by step
Paste your text into the box, or add a .txt, .docx or .pdf file. Up to 5,000 words can be checked at once; split a longer document into chapters. Then choose where to look. Wikipedia and academic papers are ticked by default. If you already suspect a particular source, tick A source I already have and give the tool a web address, a file, or the source text itself.
Press Check for plagiarism. The progress list shows each step as it happens: how many searches were run, how many articles and abstracts were compared, and anything that could not be reached. A check usually takes ten to thirty seconds, most of it waiting for Wikipedia and OpenAlex to answer.
The report opens underneath. Your text comes back with every copied run of words highlighted, one colour per source. Sentences that need attention have a dashed underline; tap one and the page jumps to the card that explains it.
What “found in the sources checked” actually means
The headline figure is the share of your words that sit inside a run of five or more words, in the same order, that also appears in one of the sources the tool compared. Five words is the threshold because shorter runs are everywhere by coincidence: “one of the most important” appears in millions of documents and proves nothing. Runs made only of small words such as “it was in the” are ignored for the same reason.
Minor edits do not hide a match. Before comparing, each word is reduced to its root, so “times” still matches “time” and “studied” still matches “study”. A sentence where you swapped a few words for synonyms will usually still show most of its wording as matched, because the sequence around those words is unchanged.
The figure is always tied to what was checked, and the report says exactly that: how many Wikipedia articles, how many academic abstracts, and whether your own source was included. A low number means nothing was found in those sources. It does not mean your text is original everywhere, and the page will not pretend that it does.
Where this checker looks, and where it cannot
Wikipedia. The tool picks your most distinctive sentences – the ones with the rarest words, numbers and names – and searches Wikipedia for each one as an exact phrase, using the MediaWiki search interface. If an exact phrase finds nothing, it tries the sentence’s key words instead. The articles that come back are downloaded in full and compared with your text word by word, so a search result on its own never counts as a match. Up to eight searches and six articles per check. The Language setting chooses which Wikipedia is searched: English, Hindi, Bengali, Spanish, French or German.
Academic papers. The same sentences are searched in OpenAlex, an open catalogue of scholarly works that includes the Crossref records publishers register for their papers. The tool compares your text with each paper’s title and abstract. It cannot see the full text of papers behind a paywall, which is where most academic writing lives, so treat this as a check of the parts of a paper people most often lift: the abstract and its definitions.
A source you add. This is the most precise check on the page. Paste the source’s text, add its file, or enter its web address and our server fetches the page for you. Every run of five matching words is found, however long the source is.
What it cannot search. Ordinary websites, blogs, news, forums and essay-sharing sites. Searching the whole web automatically needs a paid search index; services such as Copyscape or your institution’s checker have one. For this gap the report lists your most distinctive phrases with one-click exact-phrase searches on Google, Bing and DuckDuckGo. If a result shows the same words, deal with that sentence the same way as a match in the report.
Copied, closely matching, quoted: what each label tells you
Copied word for word – at least 80% of the sentence’s words match the source and most of them are identical, not just similar. Leaving it as it is presents someone else’s wording as yours.
Closely matches – at least 40% of the sentence follows the source. This is the classic result of paraphrasing too closely: the vocabulary has changed here and there but the structure is borrowed.
Quoted and cited – the sentence is in quotation marks and a citation follows it, such as an author and year in brackets or a numbered reference. That is the correct way to use someone else’s words, and the tool says so instead of scolding you. If a quotation has no citation after it, it is listed separately under the writing checks.
Every flagged sentence shows which source it matches, how many of its words matched, and the matching words underlined.
Fixing a flagged sentence without making it worse
There are only two honest fixes. The first is to quote: keep the words exactly, put them in quotation marks, and cite the source. Use this when the wording itself matters – a definition, a striking phrase, a statistic someone else produced.
The second is to paraphrase properly, which is more than changing words. Read the source, close it, and write the idea from memory in your own structure: start from a different point, join it to your own argument, split one long sentence into two or merge two into one. Then check it against the source for accidental repeats, and still cite the source, because the idea is theirs even when the words are yours. Purdue OWL’s guide to quoting, paraphrasing and summarising explains the difference well.
Suggest other words on each card looks up alternatives for the exact words that matched, using the free Datamuse word service. Use them to find your own phrasing, not as a substitute for it. Swapping words one for one inside the same sentence structure is exactly what “closely matches” catches, and it often reads worse than either the original or a genuine rewrite.
Citing what you keep
Show a citation on a flagged sentence produces an APA-style reference for the source it matched: the article title and address for Wikipedia, and the authors, year, title and DOI for an academic paper. Treat it as a starting point. Your course or publisher may require a different style, and the APA’s own citation guidance is the authority on the details.
A note on Wikipedia itself: many teachers do not accept it as a source. If a sentence matched Wikipedia, the better move is usually to follow the article’s own references to the original source, read that, and cite it instead.
Comparing a class set, or two versions of your own work
The second tab, Compare documents with each other, takes two to twelve documents – files or pasted text – and compares every one with every other. The table lists each pair with the share of the first document found in the second and the share of the second found in the first, highest first. The two figures differ when one document is much longer than the other. View opens both side by side with the shared wording marked.
Teachers use this to spot students working from the same text. Writers use it to check a new piece against their own earlier work, since reusing yourself without saying so is still a problem for many journals and course rules. This comparison runs entirely in your browser; nothing is uploaded. For a line-by-line view of what changed between two drafts, the Text Diff Checker is the better tool.
The writing checks the report adds
Alongside the source matches, the report looks inside your text for things a marker or editor notices.
Quotations with no citation. Words in quotation marks with nothing after them to say where they came from.
A paragraph that reads differently. Each paragraph of 60 words or more is measured on four things: average sentence length, average word length, share of long words, and commas per hundred words. A paragraph that sits far from the others on at least two of those is flagged. Pasted text often shows up this way, because it was written by someone else in a different register. It needs at least four such paragraphs to judge, and it is a prompt to look again, never proof.
Repeated sentences, stock phrases and very long sentences. The same sentence used twice, phrases such as “research shows” or “since the dawn of time” that invite the question “which research?”, and sentences over 35 words, which are harder to follow and harder to put in your own words. For a fuller picture of how easy your writing is to read, try the Readability Checker.
What is sent where
Your full text never leaves your browser. The comparison itself happens on your device.
To search, the tool sends a few short phrases from your text – at most eight, each nine words or fewer – directly from your browser to Wikipedia and to OpenAlex. If you ask for word suggestions, the matched words go to Datamuse. If you enter a web address as your own source, that address goes to our server, which fetches the page and returns its text to your browser. Nothing is stored by us, and no account is involved. The Privacy Policy lists these services.
If your text is confidential, untick Wikipedia and academic papers and compare only against a source you add, or use the document comparison tab. Neither sends any of your words anywhere.
Questions people ask before they submit
Is this the same as Turnitin?
No. Turnitin and similar services compare against their own databases of web pages, published papers and previously submitted student work, which no free tool can reach. This checker covers Wikipedia, academic abstracts, and any source you give it. If it flags something, your institution’s checker will almost certainly flag it too; if it finds nothing, that is encouraging but not a guarantee.
Why did a sentence I wrote myself get flagged?
Check the underlined words. A five-word run can be shared by coincidence in technical writing – a formal definition, the name of a law, a standard phrase in your field. If the run is a fixed term, it is fine to keep. If a whole clause matches, you may have absorbed the wording while reading.
My figure is 0%. Does that mean my essay is original?
It means none of the sources checked contains a run of five of your words. Use the exact-phrase search links for the open web, and remember the check cannot see paywalled papers, books or other students’ work.
Can it check a scanned PDF?
Not directly – a scan is a picture with no text in it. Run it through the Image to PDF OCR tool first, or extract the text with PDF to Text if it already has a text layer, then paste the result here.
Does it detect AI-written text?
No, and it does not claim to. It finds wording that exists in a source. Text produced by an AI model may match nothing at all and still not be your own work, so the questions of authorship and honesty remain yours.
Is it free to use for a whole class?
Yes. There is no account, no limit on the number of checks and no watermark on the report. Each check covers up to 5,000 words, and the document comparison handles up to twelve documents at a time.