Paste a draft and this reads it against itself. It finds the passages you repeated, the quotations with nobody attached to them, the citations that do not match your reference list, and the figures with no source. Four checks, marked in place, in this tab.
This check reads your document against itself. It does not search the web, any journal database, or any archive of previously submitted work, so it cannot tell you whether a passage appears somewhere else. It finds text repeated inside the document, quotations with no attribution nearby, citation problems, and figures with no source. It is not a substitute for an institutional plagiarism check such as Turnitin.
Runs of words that appear more than once in the document. A paragraph pasted twice, a thesis restated verbatim, a section padded by repetition.
Quotation marks with nothing nearby that says whose words these are. The reader can see that you are quoting and cannot see who you are quoting.
Citations that are broken, that point at a reference entry which is not there, and reference entries the text never cites.
Figures and appeals to research with no source in the same paragraph. These are the sentences a marker asks you to defend.
Everything above compares your draft against itself. This searches the live web for sentences from it, which is what a plagiarism check actually means. Distinctive phrases from your document are sent to a search provider to do it — nothing is stored, but the phrases do leave this tab, so it runs only when you ask.
No account and no word limit. The four checks above update as you type, in your browser.
The two get sold as the same product and they answer different questions. Knowing which question you have decides which tool you need, and it is usually both, in this order.
Searches a corpus — the open web, licensed publisher content, and in the institutional products, previously submitted student work — and reports the passages of your text that overlap with something in it. When it flags a line it can name the source and show you the match. That is a checkable claim, and it is the only way to answer "does this appear somewhere else".
Reads one document against itself. It finds the passages you repeated, the quotations nobody is credited for, the citations that do not resolve against your reference list, and the figures with no source. Every finding points at a specific span of your own text and explains itself in a sentence.
A similarity report is a verdict. It arrives after you submit, it is generated by someone else, and by then the defects it does not measure are already in the document. The problems on this page are the ones still under your control while the draft is open.
These are the actual thresholds the engine runs on, read out of the code that runs on your draft. Nothing here is tuned per document.
Runs of words that appear more than once in the document. A paragraph pasted twice, a thesis restated verbatim, a section padded by repetition.
Reports any run of 8 words or longer that occurs twice, matched on lowercase text with apostrophes and hyphens normalised. Every occurrence is marked.
Quotation marks with nothing nearby that says whose words these are. The reader can see that you are quoting and cannot see who you are quoting.
Audits inline quotations of 6 words or more and block quotations of 4 or more, then looks 20 words either side for a named source, a citation, a page reference, a reporting verb or a dash byline.
Citations that are broken, that point at a reference entry which is not there, and reference entries the text never cites.
Reads the document for APA, MLA and numeric citations, finds the reference list by its heading, and cross-checks the two in both directions. The cross-check only runs when a reference list is actually present.
Figures and appeals to research with no source in the same paragraph. These are the sentences a marker asks you to defend.
Flags percentages, large numbers, money, ratios and multipliers, plus phrases such as "studies show", when no citation marker appears within fifteen words inside the same paragraph. Bare years are not statistics, and the author's own results are exempt.
There is a version of this question no browser-side tool can answer, and pretending otherwise is how people get caught out. If your situation is on this list, get the institutional check.
Many universities provide one. It often sits inside the assignment page in your course system as a draft or practice submission, and where it does not, a supervisor or the library can usually run a document for you. Ask before you submit, not after.
If the assignment brief says the submission goes through Turnitin, the similarity report your marker sees is the one that counts. Nothing you run yourself replaces it.
Journals and university presses commonly run a similarity screen before peer review. Finding the match yourself is cheaper than having an editor find it after acceptance.
When there is already an allegation, the evidence is the institution's report and the sources it names. Ask to see the matched sources rather than the percentage.
Self-plagiarism is a category in many institutional policies, and only a tool holding a record of your previous submissions can see it. This check reads one document at a time.
A matcher reports overlap with other documents. Quoted material you cited correctly raises the number, and a defect that involves no other document at all raises nothing. That is why a clean similarity report and a draft with citation problems are perfectly compatible.
The absence of a corpus has one advantage worth naming. A matcher has to receive your document to compare it against anything, and the institutional products may add what they receive to the repository they search. This check has no server to send a draft to.
The four checks run as JavaScript in this tab, over a string that stays in this tab. There is no analysis endpoint behind them.
Your draft is not added to any repository, because no repository exists here. Close the tab and the text is gone.
Load the page once and disconnect. The check keeps running, because the rules came down with the page.
No. A plagiarism checker matches your text against a corpus of documents that already exist and names the ones it matched. This tool has no corpus and makes no network request. It reads your document against itself, which is a smaller job and a different one.
It indexes a large body of text — the open web, licensed publisher content, and in the institutional products, previously submitted student work — and reports the passages of your document that overlap with it, with a link to each source. The number people quote as a "similarity score" is the share of your words that matched something, which is why correctly quoted and cited material still raises it.
Because the problems it finds are the ones you can still fix, and they are invisible to a similarity report. A quotation with no attribution, a citation with no matching reference entry, a figure with no source and a paragraph you pasted twice are all defects in the draft, and a matcher will not tell you about any of them.
No. The web check searches a general web index, which is not the corpus Turnitin licenses — it has no access to student paper archives or paywalled journals. It will catch a passage lifted from an indexed web page, and it will miss one lifted from a database only Turnitin can see. Treat a clean result here as necessary, not sufficient.
The four checks above are JavaScript running in this tab — load the page once, disconnect from the network, and they keep working, because the rules are already here. The web check is different and says so on the button: it sends distinctive phrases from your document to a search provider, because searching the web is not something a browser can do by itself. It only runs when you press it.
It looks for an attribution signal within 20 words either side of the quotation. A citation further away than that, or attribution carried across a paragraph break, reads as absent. Move the citation next to the quote and the finding clears — which is usually the right edit anyway.
The cross-check reads your reference list and looks for each entry's surname and year in the body text. An entry that never appears is either a citation you dropped during editing or a source you read and did not use. Both are worth resolving before you submit.
No limit and no account. The one floor that bites is the repeated-passage check, which needs at least 16 words before a repeat is even possible.
Free, no account, and the four checks run in your browser.