AI Detection
Published June 11, 2026
Turnitin isn’t waiting at the submission gate anymore. With its Feedback Studio integration now live inside Google Classroom, the platform has moved into the space where students actually write — and that changes what “detection” means in a way most students haven’t fully processed yet.
Until now, the question was: does your finished document look AI-generated? That’s a product question. The new question is: does the way it was produced look AI-generated? That’s a process question. They are not the same thing — and the countermeasures are very different.
The integration, announced globally in May 2026 after a closed beta, embeds Turnitin’s AI-detection and plagiarism tools directly into Google Classroom for institutions running Google Workspace for Education Plus. Teachers can view AI similarity scores and originality reports without leaving their Classroom interface. Turnitin demonstrated the full suite at ISTE 2026 in Orlando.
That’s the baseline. The more consequential feature is Turnitin Clarity — a separate licensed module called Student Writing that logs the entire composition timeline. It shows when text appeared on the page, how fast, whether sections were written linearly or in chunks, and where editing happened. It is, in plain terms, a keylogger for essays.
of English submissions to Turnitin between October 2025 and February 2026 had 80%+ AI-written content — up from 3.3% in 2023.
Source: Turnitin internal data, May 2026
Those numbers explain why Turnitin is escalating. HEPI’s 2026 study found 94% of students are using generative AI on assessed work. From an integrity vendor’s perspective, a patch-and-detect approach on the finished document was never going to hold against that volume. Watching the process is the logical next move.
The complication — and it’s a real one — is that the process signal is noisy. Students use grammar tools that replace text in bulk. They write sections in Notion or Google Docs before moving them. They paste in notes, rework voice recordings, copy from their own drafts. A fast paste of 300 words is not uniquely an AI signature; it’s also how millions of careful writers work. Turnitin CEO Chris Caren acknowledged the challenge indirectly: “the rise of generative AI brings new challenges to the development of critical thinking skills” — which is a polite way of saying the signal-to-noise problem is hard.
For anyone tracking how well current bypass techniques hold up against AI detectors, Clarity introduces a new axis the tools simply don’t address yet. Humanising the text to fool a similarity score is a document-layer intervention. Clarity operates at the session layer. These are independent systems, and for now, most humaniser tools are only solving for one of them.
The practical upshot: Google Classroom just became a more monitored space than most students realise. Whether that’s appropriate is a debate worth having — process monitoring raises real questions about student privacy and cognitive autonomy that institutional procurement teams, not students, are currently deciding. For now, the best thing anyone in this space can do is understand exactly what the system can and can’t see — and go in with eyes open.
Tracking the shifting detection landscape? See our full coverage of AI detection tools, bypass rates, and what actually works in 2026.
