“I had a grandmother ask me this week whether helping with ChatGPT on the science report counts as cheating. She didn't need a philosophy. She needed a color.”We read MIT Technology Review's August 24, 2026 case study of Cheshire Academy in Connecticut, where assignments are labeled green, yellow, or red for AI use, then translated that vocabulary into a household rule you can say out loud.
When is my kid allowed to use AI on homework?
Use a traffic-light rule at the kitchen table. Green means AI is allowed. Red means none. Yellow means some tools: spell-check yes, a chatbot writing the paragraph no. You do not need a school memo to start tonight.
The hard part is not the colors. Most homework arrives unlabeled, the kid already has ChatGPT on a phone, and "helping" can mean anything from circling a misspelled word to pasting the whole prompt into a chatbot. A vibe is not a rule. A color is.
Cheshire Academy is a private boarding and day school in Connecticut, about 400 students in grades 9 through 12. On August 24, 2026, MIT Technology Review's Peter Hall reported that the school now labels assignments like traffic lights. Green: AI is fully allowed. Red: none. Yellow: the teacher can permit some tools and ban the rest, such as spell-check but not messaging a chatbot. That last distinction is the one families actually need.
Miriam Przybyla-Baum, who teaches French there and has built up classroom materials across nearly thirty years, has students let a chatbot edit their homework, then decide which edits were right and which ones stole their voice. That is a teaching exercise, not a permission slip to outsource the paragraph. Borrow the habit at home if you want. If the kid already wrote the thing, asking a chatbot to flag typos is yellow. Asking it to write the five-paragraph essay on last summer's Yellowstone trip from a blank page is green only if the teacher said so. If the point of the assignment is their own sentences, it is red, and the chatbot does not get a vote.
One honest reaction: I am tired of waiting for a perfect district policy while the report is due at 8 a.m. Schools will keep arguing. The household still has to decide tonight.
This is a household translation of one school's method as of August 2026, not a claim that your grandchild's teacher uses these labels. Check the assignment sheet, then pick a color anyway.
Where this lands in real life: The next time a grandchild texts a photo of that Yellowstone essay, due tomorrow, and asks whether they can use ChatGPT, answer with a color instead of a lecture. If the teacher never labeled it, you pick: red if the point is their own sentences, yellow if they already wrote it and want spell-check, green only if the teacher said AI is part of the work. The honest answer is now a sentence you can say before anyone opens the app.
The useful household skill is not catching the chatbot. It is agreeing, before 9 p.m., which color tonight's assignment is.
The next time a grandchild texts a photo of that Yellowstone essay, due tomorrow, answer with a color instead of a lecture: red for their own sentences, yellow for spelling help, green only if the teacher said AI is part of the work.
Ask what color this assignment is. If nobody labeled it, you pick. Red if the point is their own sentences.
Questions people ask
What if the teacher never said anything about AI?
Then the assignment is unlabeled, not unrestricted. Default to red when the work is supposed to show what the student can do alone: a quiz, a first draft, a translation they are learning to make. You can loosen to yellow for spelling and formatting once their own sentences are on the page. Silence from the school is not a green light.
Is spell-check the same as ChatGPT?
No. Yellow exists because those are different jobs. A spelling pass on a paragraph the student already wrote is closer to a dictionary. A chatbot that invents the paragraph, the examples, and the conclusion is doing the assignment. If you cannot tell which one happened, ask the kid to explain the last sentence in their own words, with the screen closed.
Should I use ChatGPT myself to help them?
Only after you pick a color for your help, not just theirs. If you want a tool that will refuse to write the answer, the homework traffic-light helper in the Vault is built for that moment. This post is the other half: the rule you say out loud before anyone types.
Related
ChatGPT for Teens is live — and some under-18 accounts are already in it
OpenAI launched ChatGPT for Teens. The useful story for families is what changed by default, and which parental controls still require setup.
Why Does AI Make Things Up, and Can I Actually Trust It?
AI tools make things up. It's called hallucination, and it's not a glitch. Here's what's actually happening and which tasks are genuinely safe.