Turnitin Clarity Tracks How Students Write, Not Just What They Wrote

A different kind of evidence

For two years, the argument over Turnitin's AI-detection score has been about accuracy: independent research, including a Stanford-led study, found some detectors misclassified non-native English writing as AI-generated at an average false-positive rate above 60 percent. Turnitin itself has warned that its own score "may not always be accurate" and should not be the sole basis for disciplinary action.

Its response, launched last year and now expanding into UK secondary schools, is Turnitin Clarity. Rather than scoring the finished text, Clarity logs the writing process: keystrokes, pasted text, deletions, time spent per section, and — where a student used Turnitin's own AI chat feature — the full prompt history. According to Turnitin's press materials, nearly 100 secondary schools and districts in the US launched pilots within the tool's first 60 days, and it was named to Time's list of the best inventions of 2025. Turnitin frames this as a shift from "gotcha" detection to transparency: teachers see how a piece of work was built, not just a percentage.

The company has publicly pushed back on characterising this as surveillance. In a letter to the Chronicle of Higher Education, Turnitin argued that Clarity "does not monitor how students are typing" in real time — it captures periodic drafts and replays the drafting process afterwards, rather than streaming every keystroke live. That is a meaningful technical distinction, but for a teacher deciding whether to trust the tool, and for a parent asked to consent to it, the practical experience is similar: a record of how a piece of writing came into being now sits on a vendor's server.

What this actually changes in the classroom

Process evidence is, in principle, more defensible than a probability score. A student who produces a rough outline, several revised drafts and visible false starts looks different from one who pastes a finished essay in one go. That distinction is intuitive to any teacher who has watched pupils work in class. Formalising it removes some of the guesswork that made the old AI score so contestable in appeals.

But it swaps one kind of workload for another. Reviewing a percentage takes seconds. Reviewing a drafting timeline — deciding whether three hours of gaps and a burst of pasted text is suspicious or just how a distracted 15-year-old writes on a Tuesday evening — takes judgement and time that departments already stretched by marking loads may not have. Turnitin's own guidance still tells teachers to combine any signal with knowledge of the student and a conversation about the work, which is sound advice but does not reduce the hours involved.

The UK data question

Turnitin Clarity's rollout has so far been concentrated in the US, but the company has signalled interest in secondary schools more broadly, and UK institutions already use Turnitin's core plagiarism and AI-detection products. Under UK GDPR, a pupil's drafting history — timestamps, revision patterns, chat logs with an AI tool — counts as personal data once it can be tied to a named student, and schools processing it need a clear lawful basis and a data protection impact assessment, not just a vendor's terms of service. Guidance aimed at school leaders has noted that data processed transiently and then discarded sits in a different category to data retained for months as evidence in a misconduct case. Clarity is built to retain exactly that kind of record, which pushes it toward the latter category and the compliance obligations that come with it.

The shift from "what score did the AI give this essay" to "what does the drafting history show" is real progress on accuracy — but it moves the burden from an algorithm to a data-retention policy, and UK schools have not yet worked out who owns that policy.

What to do

Schools considering Clarity, or products like it, should ask three questions before rollout: who can access the drafting record and for how long is it retained; what happens to a pupil's data if they leave the school mid-year; and whether the data protection impact assessment has been completed and shared with governors, not just IT. Departments should also agree, in writing, what threshold of "process irregularity" actually triggers a conversation with a student — without that, Clarity risks becoming a new source of the same disputes the old percentage score generated.

What to watch

Watch whether JCQ or the Department for Education issues any specific guidance on process-tracking tools, given that its existing position — that a detection score alone is not evidence — was written with the old generation of tools in mind. Watch too for the first UK appeal case that turns on a drafting timeline rather than a similarity score: it will set the practical bar for what counts as fair evidence in this new model.

Sources: