Before They Ask the Teacher

This week, Common Sense Media released the first census it intends to repeat every year on how American kids ages nine to seventeen use artificial intelligence. The survey reached 1,204 children. Eighty-six percent said they had used a generative AI tool. Eighty-five percent of those had used one for schoolwork. About half use it weekly. A fifth use it daily.¹ The numbers were big, and they traveled in the outlets that cover education the way headline numbers usually do.

The number to dwell on is a quieter one. Nearly a quarter of the kids in the survey said they would turn to a chatbot for help with their homework, their health, or a personal problem before they would ask what the researchers called a trusted adult, by which they meant a teacher, a school counselor, or a parent.² The finding was the lead in Education Week's coverage and the throughline of most of the rest. It is also the part of the report that touches a piece of school older than any of the tools being studied.

What Happened to the First Stop

For most of the long history of the classroom, the relationships in a child's day had an order. When the homework got hard, she raised her hand. When she could not raise her hand, she asked the kid next to her. When she got home, she asked her father. If he could not help, she went back the next morning and asked the teacher. Each of those stops was also, quietly, a piece of the record. The teacher knew what she had been stuck on. The father knew what the school was teaching that week. The friend, in the small way that friendships hold these things, knew what she was working on.

The chatbot is a different kind of first stop. It does not remember her tomorrow. The adult who tries to help her next cannot read what she asked it. The friend at the next desk has no idea she asked at all. The polite, fluent answer that lands in her notes is the only artifact of the exchange. By the time her teacher sees the homework on Wednesday, the question and the asking have already passed silently into the past.

The same census found that just over half of these kids said their school had taught them how to use AI safely, and forty-four percent had never had a conversation with a parent about it.³ The adults closest to the child are not less interested in her learning. They are simply being looped out, by no one's deliberate design, by a tool that does not loop them in.

The Quiet Rerouting

Common Sense did not frame any of this as a crisis. It framed it as a starting line. The report is the first in an annual series, modeled on the same organization's long-running census of children's media use, so the field can finally watch the relationship between kids and AI move over time rather than guessing at it.¹ That is the right posture. Nothing in the report says the chatbot is replacing the adult. It says the chatbot is, for a meaningful share of children, being asked first.

The careful question for the next decade of school is what happens at the second stop. A child who asks the chatbot at nine on a Tuesday night will, the next morning, sit in a room with someone whose job is to help her learn. The teacher will look at the homework. She will mark it. She will, if she is lucky, get sixty seconds with each child during the period to ask how it went. None of that sixty seconds will tell her what the child asked the chatbot the night before, or what the chatbot said, or what the child rejected from the answer, or what she added on her own.

That is not a small loss. The thing the chatbot has absorbed is not the answer. It is the question. The question is what teaching has always run on. A teacher who knows the questions her students are asking has the basic equipment of the job. A teacher who sees only the answers is reading the dust the practice left behind.

A Future That Holds the Question

The classrooms that will steady themselves through the next decade are not the ones that will ban the chatbot. The census makes plain that the ban, even if a school wanted one, has already lost. The classrooms that steady themselves will be the ones that find a way to keep the question visible, even when the first stop happened somewhere the school cannot see.

That has design implications a policy document cannot meet. It implies a school that has chosen, with some intentionality, which moments of practice it wants on record. The draft a student tried before she opened the tool. The prompt she wrote, and the second one she wrote after the first one disappointed her. The sentence she rejected from the answer. The line she added on her own. The minute thirty-seven breakthrough she did not know was coming when she sat down at minute thirteen. None of that is the homework. All of it is the question.

This is one of the quieter pieces of why Koan was built the way it has been built. Not as another first stop to replace the chatbot. As a way of letting the second stop, which is to say the conversation with the teacher who actually knows the child, finally have something to look at. The trail of practice. The shape of how the work came to be. The places where the child paused and the places where she did not. When that record exists, the teacher walking into Wednesday morning has the question back. When it does not, she has only the answer, and the most consequential part of the night before has happened in a room she will never enter.

Common Sense's founder called the speed of the change a wake-up call, noting that AI's arrival in childhood has happened in about three years, roughly twice as fast as social media took hold.⁴ That speed is the part that should keep the field honest. The schools that decide what they want to keep watching now will have a record to compare against next year, when the census runs again. The schools that wait will have, again, only the eventual essay.

The child still asks her question. That has not changed in any of the years school has existed. The question now is which adults in her life get to hear it.

If you could read the questions your students asked a chatbot last night, would the conversations you had with them today look any different?