November 10, 2026 · A new policy report drawing on an older survey
OpenAI’s new education report puts classroom guidance back on the agenda. For families and teachers, a practical next step is a clear agreement about what AI help is allowed on each assignment.
Read the survey carefully
The underlying ThinkYoung and OpenAI research surveyed 5,000 students aged 13–19 across ten European countries between April 14 and May 15, 2026. Of the teenagers surveyed, 79% said they had used AI for homework; 47% agreed that teachers show them how to use it. The 32-percentage-point difference between those answers does not establish that exactly 32% received no guidance.
These are self-reports, not measured learning gains or evidence of a causal effect. A separate teacher sample had a different country mix, so the broader report’s twelve-country geography should not be described as twelve student countries. The November news is a policy report, not a new survey or product launch.
A five-part homework agreement
The following is our editorial checklist, not an intervention validated by the survey. Adapt it to the assignment, school policy and the student’s age:
- Permitted help: state whether AI may explain a concept, suggest questions, review a draft or generate text. A rule that simply says “use responsibly” leaves too much unresolved.
- Disclosure: ask the student to briefly describe which tool helped and what they changed or checked. Keep this proportional to the task.
- Sources: open the original reference and verify that it supports the claim. A plausible citation is not enough.
- Privacy: avoid entering classmates’ work, personal records or identifying details into an unapproved tool. Follow school account and age requirements.
- Independent explanation: have the student explain the answer, method or revision in their own words, without the tool. This can reveal what needs further teaching.
For example, a teacher could permit hints on a practice problem while requiring the final steps to be written independently. For an assessed essay, the allowed assistance may be narrower. Write that distinction before students start rather than deciding after submission.
Revisit the agreement when the task changes. Record misunderstandings and improve the wording, rather than treating confident AI output as evidence of understanding. The co-branded research provides context for that conversation; we have not verified funding arrangements or independent validation of the survey. See the policy report and underlying research report for their respective scope and recommendations.