Students Are Already Using AI. Families Still Need to Decide How.

Two figures in this year's Stanford AI Index are worth reading together. Four out of five U.S. high school and college students now use AI for schoolwork. Yet only half of middle and high schools have an AI policy, and just six percent of teachers say those policies are clear.

The figures describe the United States, not Chinese families. But the gap they reveal is familiar: students have already started using the tools, while the adults around them are still trying to work out the rules.

In many homes, the first “AI rule” is not agreed around a table. It begins when a parent sees an unusually polished assignment and asks, “Did you write this, or did AI?”

The student immediately knows that the expected answer is not a careful account of how the tool was used. It is: “Of course I wrote it myself.” Once the conversation feels like an investigation, honest AI use starts going underground.

Families cannot wait for every school rule to be settled. We are not going back to the moment before students knew these tools existed. Nor should every use be accepted simply because AI is now common. “Allowed” and “banned” are no longer enough on their own.

Never make AI use something the child has to hide

This may sound permissive. But if admitting to AI use automatically leads to punishment, a child has no reason to describe what actually happened. Every other family rule becomes hard to apply because the parents no longer know what they are dealing with.

Instead of opening with “Did AI write this?”, try asking, “Where did AI help, and where did you decide not to follow it?” Asked that way, the student knows that using AI is something they can discuss honestly. It also asks them to notice their own decisions.

Parents can make that honesty normal by going first. You might say, “I used AI to organize this email, but it made the tone too stiff, so I rewrote the ending.” An ordinary example like that teaches two things at once: adults use AI too, and its first answer does not have to be accepted.

Not every kind of AI use is the same

A single rule for every situation will be either too loose or impossible to follow. A simple traffic-light system is easier to remember:

The exact boundary will vary by age, subject, and school. Using AI to check a practice essay is different from using it during a closed-book assessment. Asking it to explain a public research paper is different from pasting in a classmate's private messages. Where a teacher or school has set a rule, that rule comes first.

Yellow-light use deserves the longest conversation

Green-light use is usually straightforward, and red-light use calls for a clear stop. Most family disagreements happen in the middle. AI may have written part of the code, reorganized a paragraph, or suggested the sources. The work is neither untouched by AI nor automatically meaningless.

Rather than arguing about what percentage came from AI, ask the student three questions:

These questions are not a courtroom test. They are a way to make the invisible part of the work visible. A student who says, “AI wrote the function, but its first version failed on empty inputs, so I changed this condition and tested these three cases,” is showing something very different from a student who can only say, “It works.”

The same applies to writing. If AI reorganized a paragraph, can the student explain why the new order is better? If it supplied an example, did the student verify that the example was real? The important evidence is not a suspiciously exact percentage. It is the trail of decisions the student can account for.

The rules have to apply to adults too

Children notice inconsistency quickly. If an adult uses AI to write emails, summarize documents, and plan trips but treats a student's translation help as a moral failure, the family is unlikely to have an honest conversation for long.

The same three expectations can apply to everyone: say where AI helped, check important claims, and do not upload somebody else's information without permission. A parent can model those habits more convincingly than they can lecture about them.

Privacy deserves particular attention. A child's full name, school records, medical information, unpublished applications, and other people's conversations should not be pasted into a model merely because doing so is convenient. Before sharing anything, ask: would I be comfortable sending this to a stranger?

Try the agreement for one week

A family does not need a twelve-page policy. For one week, ask everyone who uses AI for a meaningful task to leave three short notes:

At the end of the week, spend ten minutes comparing notes. Which use genuinely helped? Where was AI confidently wrong? Which rule was too strict, too vague, or missing altogether? Then revise the agreement.

The tools will change, and the boundary will move with the child's age and the work in front of them. A useful family agreement should change with experience; it does not need to be perfect on the first attempt.

The goal is not to monitor every prompt. It is to help a young person use a powerful tool in the open, recognize when a line has been crossed, and gradually manage that boundary without an adult looking over their shoulder.

If a rule only teaches a child to hide their AI use more skillfully, it has probably failed.

What rules does your family currently use? Was there one you only created after something went wrong? We would genuinely like to hear what has worked, and what has not.