Parents: the try‑first homework rule
You do not have to choose between blocking AI and looking the other way. Agree one rule with your teenager: AI is allowed for homework, after you have tried, and you can show me your attempt. Then set the AI up to back the rule, and put it in writing together.
Why a rule beats a block
Blocks tend to fail in the same ways. Block one AI and another is a tab away. AI is now inside search and messaging apps. And some schools now ask students to use it. A Pew survey published in February 2026 found 54% of US teens had used chatbots for help with schoolwork.
The research points at order, not abstinence. In a trial with about 1,000 high school students, practising with an AI that gave answers led to exam scores 17% lower than for students without it, once it was taken away. A version that gave hints largely avoided that (Bastani and others, PNAS, 2025). So the useful question at home is not "did you use AI?" but "what did you try first?"
The rule, in three parts
Try first
Before asking an AI about homework, your teenager writes their own attempt: a guess, a rough answer, or exactly where they are stuck. Five minutes is plenty.
Coach, not answers
The AI is asked to check the attempt and give hints. Set this up once with the coach setup, on ChatGPT or Gemini, with your teenager sitting next to you.
Show the attempt, not the chat
If you ask, they show you the attempt. You are not reading their conversations. This keeps trust on both sides.
What ChatGPT's parental controls can and cannot do
Checked 13 November 2026 against OpenAI's help pages. These controls change often.
| They can | They cannot |
|---|---|
| Link a parent account to a teenager's account | Stop the teenager unlinking. The teenager can leave the link. |
| Set "Study hours", when new chats start in study mode (hints and questions) | Block ChatGPT during those hours. Study hours change how chats start. They do not lock anything. |
| Set quiet hours | Show you when or how long ChatGPT was used |
| Apply to ChatGPT | Cover Gemini, Copilot or any other AI |
They are worth switching on. They are not a substitute for the rule, because the rule is the only thing that travels across every AI.
Settings that support the rule
- Coach instruction: pasted once into ChatGPT or Gemini, it makes the AI ask for an attempt first. A teenager can remove it, so treat it as part of the agreement, not a lock.
- Phone limits: Screen Time or Family Link and Digital Wellbeing for homework hours.
- Computer limits: Chrome, Edge and Chromebook, with the honest list of gaps.
Ages. ChatGPT is for 13 and up, with a parent's permission under 18. Gemini for under-13s is only available through Family Link. Claude is for adults: its terms require users to be 18 or older, so it is not for teenagers.
Questions
Should I just ban AI for homework?
You can, and for younger children it may be right. For teenagers a ban is hard to enforce and easy to get around. "Allowed, but try first" is a rule they can keep, and it builds the habit they will need anyway.
How do I know they really tried first?
Ask to see the attempt. A real attempt has wrong turns in it. You are checking for effort, not for right answers.
Does Headwork monitor my child?
No. Nothing here watches, records or reports on anyone. The tools store what a person writes on their own device and send nothing to us or to you.
What does the Family plan add?
The browser extension on up to 5 browsers, with a parent PIN so the rules stay as you set them, homework hours on every AI site, and a week view you and your child can both see. Choose a code of 8 or more characters with letters, not a short number, and remember that an extension can be removed: it works beside a rule you agreed together, not instead of one. The rule, the agreement and the coach setup are free. Pricing.