What happens to text you paste into an AI detector
You are handing unpublished work to a third party. Worth two minutes of thought before pasting a thesis chapter into whatever came up first.
Every detector receives your text. What it does next varies enormously and almost none of them put it on the page you are actually looking at.
This matters more for some documents than others. A course essay is one thing. An unpublished paper, a thesis chapter, a manuscript under review or anything with a confidentiality obligation attached is another.
The questions worth asking of any tool
Is the text stored? Some tools retain submissions indefinitely. Some retain them to improve their models. Some do not store them at all. The three are very different and the difference is usually buried.
Is it used for training? A submission that becomes training data has left your control in a way that deleting an account does not undo.
Is it shared with anyone? Some tools run their detection through a third-party service, which means your text reaches a company you have never heard of.
Is there a retention period? "We delete after 30 days" and silence are not the same answer.
Do you need an account? An account ties every document you have ever checked to your identity and email address.
The submission-system case is different
Work submitted through your institution's plagiarism tooling is generally added to a repository by design, so it can be matched against future submissions. That is the intended behaviour of the similarity check rather than an accident, and it is set by your institution's own arrangements, which we are not the right people to describe.
Worth knowing that it is a separate question from what a detector you run yourself does with the same text.
Where we stand
We do not store your text. Not for training, not for a retention window, not at all.
What reaches us is a small amount of anonymous metadata: how many words the document had, the resulting score, the language, and a country code. No text, no name, no email, no account, because there is no account to have.
That is also why there is no word cap. There is nothing to meter.
What that lets you do
Run a full thesis chapter, an unpublished manuscript or a paper under review without adding a disclosure problem to whatever you were already worried about.
And run it as many times as you like, which matters because the useful workflow is check, fix the two or three paragraphs that sit highest, check again. See check your essay before you submit it.
Ours is free and unlimited with no account. It takes .pdf, .docx and .tex, scores every paragraph separately, and publishes its wrong-flag rate on genuine human writing: 0.4%, which is 99.6% specificity, against Turnitin's 4%.
