Turnitin has rolled out some big updates in 2025 to its AI detection tools — and they feel like a response to the ever more creative ways students might try to hide or “humanise” AI-generated writing.
The most headline-grabbing change? AI bypasser detection.
This is new technology designed to sniff out text that was originally written by AI, but then altered by so-called “humanizer” tools in order to slip past older detectors.
These humanizer tools have been becoming more common. Basically, they take AI-generated content and rework it so it sounds more natural, more human. The idea is that if you run the text through enough rewriting or paraphrasing tools, Turnitin (or similar systems) might not recognise it as AI-written.
Turnitin noticed this was a real problem, so they built bypasser detection right into their existing AI writing detection.
What this update means is that when Turnitin scans a paper, the “AI-generated only” category in its AI writing report now includes text that might have been run through a Turnitin bypasser. In other words, even if someone tried to humanize AI text, Turnitin is now more likely to pick it up.
Another big change that came earlier in 2025 was to the core AI-detection model. Turnitin says they’ve improved its ability to catch AI writing (they call this “improving recall”), while still trying very hard to keep the number of false positives low.
This is important, because if detection gets too aggressive, it starts flagging honest student work — that can undermine trust. But Turnitin seems to be aiming for a smart balance.
There have also been some useful usability upgrades. For example, administrators can now export AI writing detection data in a CSV file. That means schools can track trends in AI writing across classes, or look more closely at how students might be using or misusing AI.
On top of that, there’s a new feature that gives more flexible control: sub-accounts (for example, departments or sub-institutions) can now turn on AI detection even if it’s disabled at a higher institutional level.
It’s also worth noting that earlier this year, Turnitin added AI detection for Japanese-language submissions. So, it’s not just English anymore — the detector is being trained and expanded for other languages too.
Finally — and this is quite a moral point — Turnitin recently introduced AI paraphrasing detection. This helps educators spot when someone hasn’t just used AI, but has then used paraphrasing tools on top of it to try to hide that. That means Turnitin is no longer just looking for “did this come from AI?” but also “was this made more human via paraphrasing after AI?”
All of these updates together show Turnitin is really trying to keep pace with how students might try to outsmart detection tools. On the one hand, AI writing is obviously a powerful tool for learning, brainstorming, and writing; on the other, there’s a genuine risk of misuse.
By strengthening detection — especially to catch more subtle or disguised AI writing — Turnitin is reinforcing its role as a key player in maintaining academic integrity.
That said, this isn’t about “catching and punishing” students who use AI — ideally, it’s about encouraging honest conversations about how AI is used in learning.
These new tools give teachers more insight and more confidence, but it’s still up to educators and institutions to decide how they want to shape their policies. At the end of the day, Turnitin’s updates are a reminder that the AI arms race is very much real — and it’s happening in the classroom.