Families often discover AI use only when an assignment sounds unlike the child or when a teacher flags undisclosed assistance. A one-page family agreement prevents most conflicts by naming permitted help, forbidden shortcuts and what “show your work” means in an age of generated text. The rules below are written so an eight-year-old can follow them with scaffolding and a thirteen-year-old can co-author revisions.

Rule 1: know the teacher’s policy

Download or photograph written class rules about AI, tutoring and collaboration. Some teachers allow brainstorming and ban final drafting; others permit proofreading with disclosure; many policies changed mid-year. If the syllabus is silent, email a concise question before assuming.

When school rules and family values conflict, school rules govern graded work. Home practice can be more flexible, but children need to know which context they are in.

Rule 2: think before prompting

The child reads the assignment, writes initial ideas in their own notebook or doc, and attempts a plan before opening AI. That protects the moment when learning begins—the struggle to name what the question is really asking.

Permitted early uses might include clarifying vocabulary, generating practice problems with worked solutions hidden, or suggesting outline headings the child then fills with their evidence. Impermissible uses include pasting the prompt and submitting the reply.

Rule 3: do not submit invisible assistance

Any material influence—ideas, wording, code structure, feedback that rewrote sentences—is disclosed according to school rules. Changing a few words in generated text is still outsourcing authorship.

Keep a simple log for bigger projects: date, tool, what was asked, what was kept or rejected. Teachers increasingly ask for process evidence; families who track it avoid panic the night before a deadline.

Rule 4: verify claims and sources

Models hallucinate citations, dates and definitions. The child opens every linked source, checks quotations against originals, and compares important facts with reliable references. “The AI said so” is not evidence in history, science or current events.

Verification is a skill worth teaching explicitly: cross-check two sources, note confidence level, ask an adult when stakes are high.

Rule 5: protect private information

Do not upload personal data, classmates’ work, identifiable photos, school records or unpublished teacher materials without permission and an understanding of how the platform stores inputs. Family agreements should name what must never leave the device.

Rule 6: retain authorship

The child must explain, defend and revise the final work without the tool open. If they cannot reproduce the reasoning in conversation, they return to the learning stage—not to polish the submission.

Oral checks work well at dinner: “Walk me through your main argument” or “Why did you choose that method?” reveals understanding faster than scanning for AI vocabulary.

Rule 7: repair misuse

When rules break, the response includes honest disclosure to the teacher if required, redoing the learning work, and revisiting which step failed—policy ignorance, time pressure or temptation. Consequences should restore trust, not only punish.

Review the agreement at term start, after district policy updates, and after any misuse or confusion. Tools rename themselves monthly; behavior-based rules survive platform churn.

How rules change from ages 8 to 14

Children around eight to ten often need adults to control access, sit nearby during first uses, and offer limited choices (“You may ask for synonyms, not for the paragraph”). By eleven or twelve, they can help design the agreement, estimate time, and discuss consequences. Early teenagers need increasing privacy for drafting, but privacy does not mean unbounded tool use—ownership includes documenting help and meeting deadlines without parental rescue.

Common parent traps

• Replacing a vague ban with a vague blessing—“just don’t cheat”—without examples.
• Using AI to finish work when the family is exhausted, then expecting the child to unlearn the shortcut.
• Confusing polished output with mastery because submission stress dropped.
• Skipping verification because the answer “looks right.”
• Treating resistance as moral failure rather than unclear rules or missing skills.

A four-week review

Choose one outcome—disclosure compliance, verification habit, or reduced last-minute panic—and one visible measure. Review weekly without turning the child into a surveillance project. At four weeks, continue, adapt or simplify. Write down the reason so the family does not revert to chaos when the next app launches.

Frequently asked questions

Review note

This guide is educational, not legal or mental-health advice. School policies, platform terms and academic integrity rules vary. Persistent changes in learning, mood or daily functioning should be discussed with qualified professionals familiar with your child.