I believe at least half of my students are using AI inappropriately as of this fall 2026 semester. They are likely violating academic integrity policies and, more importantly, they are impairing their learning. As I wrote last summer, though, the biggest damage will be to the value and practice of honesty. Perhaps higher education will evolve, but in the meantime deceit is becoming common.

For those (demoralized) instructors attempting to hold the line—I’m lucky to have to do so for only one more year—I’ve learned something about AI chat logs. I don’t ban AI, but I require student work (understanding and prose) to be substantively their own. In short, “query, don’t copy & paste.” When students are confounded as to why their prose is flagged as AI, I sometimes ask them to share the thread, or students volunteer one, to help figure it out. I’ve now seen a few cases where students claim they can’t share a thread or offer a fabricated log as evidence that the work was their own. I’ve seen:

1. Students fabricate a chat log and place it in a PDF or Word document. I respond by asking for the original.
2. Students tell me the service they used doesn’t permit a thread to be shared. I respond that this is not true.
3. After the fact, students create a new thread favorable to them, showing them asking for but rejecting the chatbot’s prose. I check the timestamps.

Yes, chatbots from the major services do have timestamps, though you might need to ask a chatbot to find the relevant attributes and convert the timestamps (from Unix epoch time).