Students, educators and people writing at work all reach for an AI pattern checker for different reasons. Here is what each of them does with it, and where it is the wrong tool.
If you are worried an essay will be read as machine-written, paste the draft in and look at what comes back. The passages that follow AI patterns are marked where they sit, each with the reason, so you have a particular paragraph to look at and not a number to worry about. Read the reason. If the paragraph is formal because the subject needs it to be, leave it. If it sounds like nobody in particular wrote it, say it again your way.
Plenty of people use AI to brainstorm, to tidy wording or to work out a structure, and that is ordinary now. The ideas and the research are yours. What tends to give it away is the phrasing a model left behind, and the grey marks show you where that phrasing is, so you can write those parts the way you would have written them. Do check what your course allows first.
An EPQ or a dissertation rarely fits into one pass. Run each chapter on its own and every result stays in Documents under its own title, which is what makes the pile navigable three weeks later. In deadline week the chapter you already checked is still sitting there.
Read your gold passages together. They are the ones that sound like you, and after a few drafts you start to recognise the habits in them: how long your sentences run, how you get into a point. That is more use than any filter, because it tells you what to write towards.
Paste in a submission and you see where a student sounds like themselves and where the writing goes flat, with the reasoning behind every mark. Start with the gold. "This section sounds the most like you, and it is the strongest part" is a better first sentence than an accusation. The marks cannot tell you who wrote what, and they are not meant to.
Any finished analysis downloads as a dated A4 PDF with the marked passages and the reason for each. It is a record of what was noticed, not a verdict on anybody. Check what your institution allows in a formal process before you rely on it.
The Educator plan carries 1,300 CVT a month, which is about twenty-three analyses of 1,100 words. Queue a stack to run overnight and each one costs about 40% less and is waiting by morning. If a month runs long the extra is metered, so marking is never cut off partway through.
Clients and employers have got good at noticing copy that sounds like everybody else's. Paste in a report, a proposal or a cover letter, and the paragraph that reads that way shows up while there is still time to say it properly.
Agencies, editors and content teams run it over submitted copy before it goes out. A marked passage with a reason attached is something you can point at in a conversation with the writer, which is easier than "this doesn't sound right".
You can run a covering letter through it, and it will show you which parts sound like a template. Please do not reject anybody on a mark. Careful writers and people writing in a second language pick up grey for nothing worse than being formal, and a mark says how a passage reads, never who wrote it.
Every account has a referral link in the account menu. When somebody joins through it they get 10 CVT, you get 10 once they have come back on another day, and if they later subscribe you get 90 more. It works on the free tier, and nobody has to raise a purchase order.
UnTXT generates nothing. It reads a draft you have already written, which is a different thing from the tools most university AI policies were written about. Policies still differ, and yours is the one that counts. We have written up what several UK universities publish, with the source and the date we checked: Loughborough, UCL, Imperial, Cambridge, Oxford, Manchester, Edinburgh, LSE and King's.
They do not all say yes. Lancaster's regulations tell students not to upload assessed work to a tool like this without permission in advance, and we have written up fourteen more universities for signed-in readers, Lancaster among them. Check yours before you check your essay.
Our advice is the same everywhere: say you used it. "I used UnTXT, an AI pattern checker, to check my own draft. It made no changes to the text."
Two AI models, Claude Opus 5.5 and GPT-6 Sol, read your text separately and argue over it passage by passage. Only what both land on gets marked, and a third model, Claude Sonnet 5, reads whatever they cannot settle. You can read what each of them said. A run of 1,100 words is roughly 55 CVT, or about 32 overnight, and the estimate is shown before you confirm.
Two careful readers agreeing is worth more than one opinion. It is still a reading of how the text sounds and not a legal finding, and Sataklela OÜ accepts no liability for decisions taken on the strength of it.