Cheshire Academy in Connecticut is a private boarding and day school serving roughly 400 students in grades 9–12. The school does not mandate AI use for instructors, but George Aiello, the school librarian and technology coordinator, says the “vast majority” of teachers use AI tools in some way.
Background
The arrival of chatbots a few years ago surprised many schools: students suddenly had apps capable of answering many homework questions or producing essays in seconds. Teachers can often detect AI-generated work because models make errors that humans typically do not, and generated text can show stylistic giveaways.
On the advice of consultants, Cheshire Academy chose not to prescribe a specific technology. Instead, the school trained staff on general AI techniques, covering how to craft effective prompts while emphasizing the technology’s limitations, including its tendency to produce incorrect or biased outputs.
Practices in the school
-
Lesson preparation and administration: Many teachers use generative AI to help plan lessons, create grading rubrics and prepare class materials. Some would like to deploy AI to help provide student feedback, but concerns about quality, personalization and privacy have so far prevented widespread use for that purpose.
-
Modeling and pedagogy: Some teachers, such as French teacher Miriam Przybyla-Baum, do not use AI personally but incorporate reflection about AI into their teaching. Przybyla-Baum has nearly 30 years of teaching experience and had seen students use translation tools and shortcuts well before ChatGPT's release.
-
Assignments that build reflection: In one exercise, students allow a large language model (LLM) to edit their homework, then review the edits to decide which were correct and which removed their voice. In another, students anonymously grade each other’s AI-assisted work, annotating parts they believe were generated by a machine.
-
Traffic-light labeling: Assignments are labeled like traffic lights: green means AI is fully permitted, red bans all AI use, and yellow allows the teacher to permit some tools while banning others (for example, allowing spell-check but not a chatbot).
-
Student involvement: The school is piloting a “Student AI Council” in which students create media and lead discussions about healthy AI use, encouraging peers to reflect on how AI should and should not be used for the community’s benefit.
Tools in use: general chatbots and MagicSchool
Cheshire Academy combines general-purpose chatbots (such as ChatGPT and Perplexity) with education-specific platforms like MagicSchool. Staff say MagicSchool’s primary strength is the breadth of capabilities in one package: it can generate quizzes, worksheets and other question sets across subjects and grade levels; it includes a grading-rubric generator that outputs ready-made tables; and it can create presentations, lesson plans and administrative reports.
Teachers input specific parameters for each task (grade level, number and type of questions, alignment documents) through a single platform. Not every educator is comfortable using LLMs to produce student-facing text — worries about accuracy and pedagogical fit remain. MagicSchool offers free and paid tiers; for unlimited access and complete record-keeping in the system, an individual plan costs just under $100 per year. Many teachers also use general-purpose chatbots from vendors like Anthropic, Google and OpenAI for administrative tasks, and some companies are rolling out school-specific features with mixed results.
Practical takeaways for applying AI in schools
-
Meet students where they are: Students will be tempted to use AI and it is difficult to police entirely for take-home work. Counter misinformation about AI’s capabilities and teach healthy strategies for interacting with the tools.
-
Model best practices: AI can automate many tasks but struggles with precision and voice. Teachers should demonstrate careful use to avoid normalizing lazy practices that students might replicate.
-
Refine and replace carefully: AI is useful for brainstorming lesson plans and extra problem sets, especially for early-career teachers without a large problem bank. Because models can hallucinate, verify every citation, equation and factual claim before presenting AI-generated material to students.
Cheshire Academy’s piecemeal, teacher-centered approach — staff training, student reflection tasks, traffic-light assignment rules and a student council — illustrates one way schools can integrate generative AI while addressing pedagogical, privacy and accuracy concerns.
