The Line They Asked For

On June 9, Oxford University Press published research drawn from nearly four thousand UK teenagers, aged 13 to 18, on how they actually use AI in their schoolwork.¹ The headline that ran in most papers was about cheating. Only 44 percent of the students believe it is cheating to use AI to complete all of a homework assignment. Eighteen percent think it is cheating to simply ask AI for tips.¹ The boundary, they confessed, is grey.

The more useful finding sits one paragraph down. Only 15 percent of the students said they have been given enough guidance by their schools on how to use AI well.¹ Forty-eight percent said they want their teachers' help in deciding what AI content is trustworthy. Fifty-one percent said they want clearer rules on when AI should and should not be reached for.² Eight in ten of them already use it.²

Read those numbers in the order the students would have meant them, and the story is not that adolescents are confused. The story is that they are asking adults to draw a line, and the adults have not been able to.

The Question Underneath the Survey

A reporter will describe the finding as a gap in policy. It is also a gap in evidence. Drawing a line about when a student should use AI requires knowing what the student is doing. For a century, schools have answered the question of what the student did by reading the artifact at the end. The homework. The essay. The exam booklet. The artifact testified, because the path between the child's thinking and the page was direct and unforgiving.

That is no longer true. When eight in ten teenagers are pasting prompts in and out of a chat window before they hand anything in, the final document is no longer a record of work. It is a record of revision, of negotiation, of someone in the room who did not sign the paper. The teacher who tries to write a guidance memo for her class is, in effect, drawing a line for behavior she cannot observe.

This is why the 15 percent number is the one to dwell on. The students are not refusing rules. They are asking for them, and the adults cannot reasonably write them yet.

What They Said They Wanted

The same OUP study asked the teenagers what kind of AI tool they would actually want in the room. Forty-four percent said they preferred tools that suggested activities to deepen their understanding. Forty-one percent preferred tools that asked them questions to help them arrive at the answer themselves. About one in five wanted an AI that simply gave them the answer.¹

Eight in ten of the students, in other words, were describing a tutor. Not a writer. Not a worksheet engine. A patient older sibling who asks the next useful question and waits.

That preference is worth pausing on. The adolescents using AI most heavily are also the ones most able to articulate what good help looks like. The line they are asking for is not a wall. It is a definition. They want adults to say: this part stays yours, and the tool exists to make it more yours, not less.

What the Senate Heard the Same Week

On June 16, the U.S. Senate Health, Education, Labor and Pensions Committee held its first 2026 hearing on K-12 AI.³ Delaware's Secretary of Education, Cynthia Marten, opened with a student's own words. AI, the student had written, should be "a collaborator and tool, not something to do all the work for us."⁴ Marten then turned to the room and said the obvious thing. The tool, she said, is no good if the person using it does not know how to use it.⁴

It is the OUP finding in a different voice. The students want a line. The teachers need a way to see whether the line is being walked.

How a School Would Begin to Answer

You can describe the answer in one sentence. You make the practice visible.

That is a small sentence with a long tail. It means letting the page record the pauses, the revisions, the rewrites, the moments a student copied something in and then took it out, the eight minutes she spent on a paragraph before her own sentence finally arrived. It means a teacher being able to read a finished essay and also read the half hour that produced it. It means that when a student asks her teacher whether something counts as cheating, the teacher can answer from the actual record rather than from a hunch.

The 15 percent is not really a guidance problem. It is a sight problem. Schools that learn to see the practice will be able to draw the line the teenagers are asking for, and to enforce it without suspicion. Schools that grade only the artifact will keep writing memos that cannot be honored, because nobody on either side of the page can say with confidence what they describe.

A Quieter Reading

The simplest reading of the OUP survey is the one most worth keeping. The teenagers are not the antagonists in this story. They are, in surprising numbers, the ones asking the adults to be clearer. They want to be told where their thinking ends and the machine's begins. They want adults who can tell. They have raised their hands.

The room, for the first time in a long while, is waiting on us.

What would it cost a school to be able to answer them honestly?