No. NoRedInk does not have an AI detector, and it does not produce an AI probability score.
That is the short answer, and it is worth resisting the temptation to add a "but" to it. NoRedInk has not built a classifier that reads a paragraph and estimates whether a model wrote it. What it has built is something different, and in some ways more interesting: a set of reports about how a piece of writing was produced. If you are a student wondering whether NoRedInk is quietly scoring your work for AI, it is not. If you are a teacher wondering whether it will tell you who used ChatGPT, it will not do that either, though it may show you something more useful.
Here is what it actually does.
What NoRedInk is
NoRedInk is an English Language Arts platform used in schools, built around grammar and writing practice, guided drafting activities and teacher feedback. Its recent development has been squarely in the AI-assistance direction rather than the AI-policing one. The Grading Assistant evaluates student writing against rubric items and suggests scores and written feedback, which NoRedInk says reduces grading time by over 40%. Notably, students never see that feedback until their teacher approves it.
But grading is not detection. NoRedInk states plainly that the Grading Assistant "doesn't provide feedback on whether students submitted original work." It is a tool for getting feedback to more students faster, not for adjudicating authorship.
What Originality Insights actually reports
The feature people are usually thinking of when they ask this question is Originality Insights, and it is a process report rather than a text classifier. NoRedInk's own materials describe it as providing visibility into the student drafting process, and its academy course calls it a "Pace and Paste Tracker", which is a fair summary of the two signals involved:
- Time spent per draft. How long the student actually spent working in the assignment. NoRedInk notes that this is measured from the point the student interacts with the assignment after completing the interactive tutorial, and that the tutorial itself is not included in the time calculation.
- Copy-paste detection, and the percentage of pasted text. Where text was pasted in from an external source, and how much of the submission that accounts for.
That is the whole mechanism. No perplexity, no burstiness, no model fingerprint, no percentage estimating machine authorship. Teachers see how the work was made, not a verdict about what made it.
What NoRedInk says it cannot do
This is the part that deserves credit, because it is more careful than much of the AI-detection market manages. NoRedInk's documentation states that Originality Insights "do not detect plagiarism." The framing is that the insights help teachers make educated decisions about whether a student has plagiarised or used generative AI. The insights are input to a judgement a human makes, explicitly not the judgement itself.
Compare that with the tools that return a confident percentage and leave the interpretation to whoever reads it, and you can see why the restraint matters. A vendor that declines to claim a verdict cannot have its verdict misused.
What a paste event does and does not prove
If you are reading an Originality Insights report, this is the important nuance: a large paste is ambiguous, not damning.
A submission that appears in one or two paste events is equally consistent with a student who:
- drafted in Word or Google Docs offline and pasted the finished piece in
- wrote on their phone and transferred it
- typed up handwritten notes
- pasted from a language model
All four produce the same trace. Similarly, a short drafting time can mean a student worked out their argument elsewhere, or that they are simply a fast writer, or that they had already planned the piece in class. Process data narrows the question and tells you where to ask it. It does not answer it.
The same holds in the student's favour, and this is worth knowing if you are the one being asked about your work. A long drafting time with incremental edits and no large pastes is strong evidence of authorship, and it is the kind of evidence a detector score can never give you. Keep it.
Why process signals beat a score, and where they stop
There is a real argument that NoRedInk has chosen the better instrument, and it is the same argument Anthology made when it tested a market-leading AI detector for Blackboard, found the error rates too high and the models biased, and published its decision not to build one.
Statistical AI detection is inference from finished text, and it is wrong a measurable share of the time in a way that falls unevenly. Liang et al. (2023), published in Patterns, ran seven detectors over TOEFL essays by non-native English speakers and found they flagged more than 61% on average, against near-zero false positives on essays by native-English US eighth-graders. Process data carries none of that bias, because it does not care what your prose looks like. It reports what the editor recorded.
The limits are just as structural, though:
- It needs the work to happen inside the platform. Anything drafted elsewhere reduces to a single paste event.
- It needs the record to exist. Retrospective work, or work submitted before the feature was in use, has nothing to replay.
- It is trivially avoided by anyone who knows about it. Retyping generated text by hand defeats paste detection entirely and inflates the drafting time at the same time.
That last point is the honest limitation of every process-based approach. It catches the careless, not the determined, which is a reasonable thing for a school platform to do but not a security guarantee.
What this means if you are a student
Nothing you need to panic about. NoRedInk is not scoring your paragraphs for AI. What it can show your teacher is how long you worked and whether text arrived by paste.
The practical consequence is straightforward: if you draft somewhere else, expect the paste to be visible, and be ready to explain it honestly. Drafting in Word is not misconduct, and saying "I wrote it in Google Docs, here is the version history" resolves the question completely. Keeping that history is the single most useful habit here.
What this means if you are a teacher
Originality Insights is a conversation starter with better foundations than a detector percentage, and it should be treated the same way: as the thing that makes you look, never as the finding.
The most valuable pattern it can show you is the combination. A long, incremental drafting history is reassurance. A single large paste is a question worth asking, gently, because the innocent explanations outnumber the guilty one. And if you also run a text-based detector, note that clean process data alongside a high AI score most likely indicates a false positive, which is exactly the situation where a case should stop before it starts.
If you need to check work that never passed through NoRedInk, such as a file submission or a document you were simply handed, process data has nothing to read and a text signal is what remains. That is a different tool for a different situation, and it comes with the accuracy limits above rather than without them.
Frequently asked questions
Does NoRedInk have an AI detector?
No. NoRedInk does not produce an AI probability score or classify text as machine-written. Its Originality Insights reports drafting time, copy-paste detection and the percentage of pasted text, and its Grading Assistant evaluates writing against rubric items. NoRedInk states the Grading Assistant does not provide feedback on whether students submitted original work.
Can NoRedInk tell if I used ChatGPT?
Not directly. It has no way to identify generated text. It can show your teacher that text was pasted into the assignment and how long you spent drafting, and a teacher may draw conclusions from that. Pasting is not evidence of AI use on its own, because drafting offline in Word or Google Docs produces the same trace.
What is NoRedInk Originality Insights?
A process report rather than a detector. It gives teachers visibility into the drafting process through two signals: time spent per draft, measured from when the student interacts with the assignment after the interactive tutorial, and copy-paste detection including the percentage of pasted text. NoRedInk's documentation states the insights do not detect plagiarism and are intended to inform a teacher's judgement.
Does NoRedInk detect plagiarism?
NoRedInk's own documentation says Originality Insights do not detect plagiarism. The copy-paste detection can alert a teacher that text came from outside the assignment, which is useful information, but it is not a similarity check against published sources and it does not identify a source.
Will NoRedInk flag me if I wrote my essay in Google Docs first?
It will show that the text was pasted, since that is what happened. That is not misconduct, and it is easily explained. Keep your Google Docs version history: a document composed over several sessions with revisions and false starts is strong evidence you wrote it, and considerably better evidence than any detector score.
Is process data better than an AI detector?
For what it covers, it is more direct, since it reports what the editor recorded rather than inferring from statistics, and it carries none of the bias against second-language English that detectors do. It stops where the record stops: work drafted elsewhere, retrospective work, and anything submitted as a finished file. It is also avoidable by retyping text by hand.
Sources: NoRedInk product and help centre documentation and its academy course materials, together with NoRedInk's own blog on the Grading Assistant, checked 29 September 2026. NoRedInk's help centre returns errors to automated requests, so quoted phrasing was taken from its own published documentation as surfaced rather than from a direct page read; treat NoRedInk's current documentation as authoritative if it has since changed. Liang et al., Patterns (2023). Anthology's published position on AI detection for Blackboard.
Try it on your own writing