How to encourage smarter AI use in the classroom

Chatbots landed in classrooms fast, and they brought a mess with them. Students suddenly had an app that could answer homework questions or crank out an essay in seconds. Teachers could spot some of it — AI text has its own tells, from obvious mistakes to a suspicious abundance of em dashes — but that didn’t make the workload lighter. It made it heavier. Lesson planning, assignments, grading, and now figuring out how to live with the new tool all sit on the same overworked desk.

At Cheshire Academy, a private boarding and day school in Connecticut with about 400 students in grades 9 through 12, the response has been practical rather than purist. Administrators don’t require teachers to use AI, but the school’s librarian and technology coordinator, George Aiello, says the “vast majority” do in some form. Some use ChatGPT or Perplexity. Others reach for MagicSchool, a platform built for educators. The common thread is that the school trained staff on general AI skills instead of forcing a single product on everyone.

That training focused on prompting, but it also put limits front and center. Teachers were told to expect incorrect and biased output, not polished wisdom. In practice, many now use generative AI for the unglamorous stuff: lesson planning, grading rubrics, and other class materials. A few want help with student feedback too, but quality, personalization, and privacy have kept that idea on the shelf for now.

Miriam Przybyla-Baum, who teaches French, takes a different route. She doesn’t need AI for her own prep, after nearly 30 years of building materials, but she does use it as a teaching subject. She had already seen students leaning on tools like Google Translate before ChatGPT arrived. Now her students let an LLM edit their homework, then comb through the changes to decide what was helpful and what erased their voice. In another exercise, they anonymously grade one another’s AI-assisted work and mark the parts they think came from a machine.

The school is pushing that idea further with a Student AI Council, where students create media and lead discussions about healthy AI use. Cheshire Academy has also adopted traffic-light labels for assignments: green means AI is allowed, red means none at all, and yellow leaves room for limited tools like spell-check while banning chatbot use. It’s a simple system, which is probably why it works. A classroom doesn’t need a grand theory of AI. It needs rules students can actually understand.