To schedule an appointment with the Coordinator of the Center for Teaching and Learning
AI detectors available online often do not offer clear information about the use of data, so we tend not to encourage their use. For ease of use, and to get familiar with similar tools, those integrated within wider platforms (such as Grammarly or Quillbot) as they should--stressing the conditional--be more attentive in terms of personal data.
Another issue arising with detectors available online is that they are increasingly requiring to purchase a licence or a subscription for more than a very limited access, so we cannot offer any proper assessment of such tools.
For the same reason, John Cabot University does not endorse any specific AI-detector per se; currently, JCU faculty members have access to Turnitin Originality, which offers a set of detecting features, such as for AI-generated content and plagiarism, within the broader context of academic integrity.
As a further strategy for AI detection, we encourage faculty to obtain diagnostic sample from the main AI models in advance. That involves running the prompt of the assignment through AI's like ChatGPT, Gemini, Claude, to see what a fully AI-generated result would look like. Of course, prompts can be fine-tuned further (e.g., "write it as a non-native English speaker").
Another option, in the case of promptless/free assignment, is to have the students themselves provide you with the prompt afterwards, as a sort of reflective practice, and to use that to run an "after the fact" diagnostic sample.
At any rate, AI detection tools and strategies, though useful, cannot and are not supposed to replace a conversation with the student.