Four ways families actually share ChatGPT — pick one for tonight

A lot of parenting advice about AI sounds like a rulebook written for someone else’s house. Ban it. Track every chat. Or shrug and hope the platform’s age gate does the rest.

Researchers at the University of Wisconsin–Madison sat with families who were already sharing ChatGPT and asked what they actually do. Shirley Zhang and colleagues interviewed 12 U.S. families — at least one parent and a child about ages 8–18 who used a shared account. The study is qualitative: small on purpose, rich on detail. It is not a percent chart. Read it here: arXiv abstract and the full HTML version.

What they found is useful at the kitchen table: families lean on a short menu of habits — and many GenAI platforms still give parents few child-specific controls, so the habits matter.

The four habits, in plain words

The team adapted classic media-mediation ideas from Nikken and Jansz into four GenAI strategies. Kitchen English:

1. Instructive — You talk about rules, risks, and what “good use” looks like. Not a lecture every night. A real conversation when the tool shows up. 2. Restrictive — Access is limited. Often: ask before you open it. Permission is the gate, not the password alone. 3. Co-use — You try prompts together, or you share what the bot spat out and talk about it. The account is a joint activity, not a secret tablet. 4. Supervision — You keep an eye on use: shared device, screen the prompt, notice the history. Not spyware theatre — proximity and shared screens.

Some families in the study mixed these. A few enforced almost nothing. The point for parents is not to copy a lab profile. It is to name the habit you already half-do — or the one you want this week.

Why “pick one” beats a rulebook

When every risk gets a new policy, kids hear noise. One clear line is easier to keep:

• “Ask before you open it.”
• “We try the first prompt together.”
• “Shared laptop only — not in the bedroom.”
• “If it sounds sure, we check one other place.”

That last one is instructive plus a tiny check. It still counts as a family rule you can say in one breath.

Fathers, grandparents, and other caregivers can use the same sentence. You do not need a specialist vocabulary. You need one line that survives Tuesday homework and Saturday boredom.

The Wisconsin–Madison interviews also remind us why this is not “set it and forget it” on the platform: many GenAI tools were not built with child-specific parental controls the way streaming apps are. If the product does little, the household has to do something small and repeatable.

What the study is — and is not

It is a careful look at how 12 families who already share ChatGPT talk about risks, opportunities, and everyday mediation. It proposes a Family–GenAI mediation model grounded in older media research.

It is not a national survey. Do not turn “12 families” into “most parents.” Do not invent a percentage for how common each habit is. The value is the vocabulary: four named strategies you can choose from without pretending science handed you a ranking.

If your house looks different — younger kids, no shared account, school-only tools — you can still borrow the menu. Restrictive might mean “not yet.” Co-use might mean “we brainstorm dinner with it on the fridge screen.” Supervision might mean the living-room laptop stays in the living room.

Tonight’s tiny ask

Before bed, pick one strategy for this week. Say it once, out loud, to the people who need to hear it.

Examples that match the research categories:

• Instructive: “If you use ChatGPT for homework, tell me what you asked.”
• Restrictive: “Ask before you open it.”
• Co-use: “We try the prompt together first.”
• Supervision: “Shared device. I’ll glance at the thread.”

Then stop. One line beats a manifesto. Tomorrow you can refine it. Tonight you only need a shared sentence.

Same love. Different tools. The platform may not parent for you. A short, spoken habit still can.