Educator quick-start
This is the short version for someone who has been handed a stack of submissions and a suspicion about one of them. If you want the click-by-click workflow, that is at Reviewing student work. This page is about what the tool is for and where its edges are.
What you get back
Paste a submission and the parts of it worth a second read come back marked, either gold or grey. Gold means the passage carries the marks of a person writing. Grey means it reads the way machine writing tends to read. Click any of them and you get the reasoning that put it there, in words rather than in a score.
There is no percentage. We took the last one out in version 4.49.0, because a number is a thing people argue about in meetings and a marked passage is a thing you can read together with the student in front of you.
What it is not
It is not proof, and we would rather be the ones telling you that.
Detection is probabilistic. Human writing genuinely does read as machine writing sometimes, and it does so most often in exactly the students you would least want to be wrong about — someone writing in their second language, someone who has been taught to write formally and does it well, someone who over-polished a draft the night before because they were anxious about it. A mark is evidence that a passage reads a certain way. It is not evidence about who wrote it, and nothing in the product will ever claim otherwise.
So the tool belongs at the start of a conversation rather than at the end of one. It tells you where to look, and then you do the part that requires knowing the student.
Why the gold marks matter more than you would think
Most detection tools only look for risk, which means the best possible outcome for a student is that nothing was found. That is a miserable shape for a conversation to have.
Marking the passages that read as genuinely theirs gives you something to open with, and it gives a student who did write the essay something to point at. In practice this changes the tone of the meeting more than any other single thing about the tool, because you can start with what is clearly their own work and ask about the rest from there rather than arriving with an accusation and asking them to disprove it.
Fewer marks on formal writing
Teacher and Professional accounts run on a stricter threshold, which we call Confidence Guard. A passage has to carry a stronger signal before it gets marked at all.
This exists because the register students are taught to write in — hedged, impersonal, carefully structured — is also the register that trips naive detectors, and a page filled with marks tells you nothing you can use. Student accounts run on the standard threshold instead, which is deliberately more generous, since a student editing their own draft loses nothing by being shown a passage that turns out to be fine.
Where educators find it useful
Mostly not in misconduct cases. The common uses are formative: showing a student what over-polishing does to their own writing, supervising a dissertation across drafts, or handing the tool to a class before submission so that checking their own work is a normal part of finishing it rather than something that happens to them afterwards.
Several teachers tell students outright that they use it, which removes the surprise and turns it into a shared standard rather than surveillance. That tends to work better than not saying.
Before you use it in a formal process
Check your institution's policy on detection tools, because it varies a great deal and some institutions have a settled position on what may and may not be cited in a hearing.
If you need a record, the toolbar above the editor has a Download, which gives you a dated A4 PDF with the word count and the marked passages. Keep it as one input alongside the student's own account of how they wrote the piece and whatever you know of their previous work, which is almost always the stronger evidence anyway.
Related
- Reviewing student work — the workflow, step by step
- False positives — who gets wrongly marked, and why
- Understanding your results — what the two marks mean
- Why percentage scores fail — the case against the number