OpenAI report finds 79% of students use AI for homework, but only 47% were shown how

OpenAI published a report, Closing the AI guidance gap in education, on November 10, signed by Tom Duff Gordon, VP of Global Affairs EMEA, and Leah Belsky, VP of Education. It draws early lessons from the company’s education partnerships across EMEA, including work with governments and school systems in Estonia, Slovakia, Greece, and Italy.

The sharpest evidence comes from new ThinkYoung research cited in the report: 79% of surveyed students reported using AI for homework or assignments, but only 47% said anyone had shown them how. OpenAI reads that gap as urgent, arguing governments need to support educators and give students clearer expectations.

The partnerships share three priorities: equitable access to frontier AI tools, jointly designed AI literacy programs, and ongoing measurement and evaluation, adapted to local languages, curricula, and institutions.

Five themes organize the lessons: clear educational goals, educator preparation, local partnerships, critical AI literacy, and credible evidence. Country experience in Estonia, Greece, and Slovakia points to strong public leadership, investment in teachers’ skills and confidence, and practical support tailored to local curricula.

On outcomes, OpenAI stays careful. Teacher adoption and reported time savings are described as encouraging early signals, while whether AI strengthens knowledge, independent thinking, and retention still requires sustained evaluation; surveys, program monitoring, and consultation are used to understand participants’ experiences and barriers.

The report also leans on a familiar argument: as AI becomes more capable, the teacher’s role becomes more important, not less, a point echoed in debates around PISA and AI including commentary from the OECD’s Andreas Schleicher.

OpenAI says the OpenAI Learning Lab is bringing education researchers together to understand when and how AI supports education systems. It frames the report as early experience rather than a single model for others to copy.

What to watch is the evidence base. Adoption rates and self-reported time savings are easy to collect; comparable learning outcomes are not, and schools are writing AI rules faster than many teachers are being trained to use the tools.