79% of Students Use AI for Homework, Only 47% Get Guidance: OpenAI Flags Education 'Guidance Gap'

By James Calloway

Seventy-nine percent of surveyed students report using AI for homework or assignments, but only 47% say anyone has shown them how. That 32-point gap is the central figure in a new report from OpenAI on its education partnerships across Europe, the Middle East and Africa, and the company frames it as an urgent problem for governments, not just schools.

The report, titled "Closing the AI guidance gap in education" and published in November 2026, was authored by Tom Duff Gordon, VP of Global Affairs EMEA, and Leah Belsky, VP of Education at OpenAI. It draws on new research from the think tank ThinkYoung, which examined how teachers and students are already engaging with AI, alongside lessons from OpenAI's government partnerships in Estonia, Slovakia, Greece and Italy, among others.

The stakes are straightforward. AI is forcing a reassessment of what education systems should prepare people for, the authors argue. As intelligence becomes easier to access and apply, the question shifts from what people need to know to which skills will help them thrive in an AI-enabled world.

Why does the teacher's role grow rather than shrink?

The report takes a position that runs against the common assumption that AI tools will sideline educators. As AI becomes more capable, the role of teachers becomes more important, not less, the authors write. Teacher expertise, judgement and ability to guide learning remain essential.

This argument aligns with recent debates around PISA (the OECD's Programme for International Student Assessment) and AI, including discussions led by the OECD's Andreas Schleicher. Those debates point to the same lesson: equipping teachers, schools and governments with AI skills and the confidence to make informed decisions about these tools must be a priority for education systems.

The implication is a policy demand, not a product pitch. Governments need to support educators and give students clearer expectations about when and how AI use is acceptable, the report states.

What is OpenAI actually doing with governments?

Across EMEA, OpenAI says it is working with governments and education systems in Estonia, Slovakia, Greece and Italy, among others, to explore how AI can support educators. The company describes three shared priorities across these collaborations:

• Equitable access to frontier AI tools (the most advanced models the company offers)
• Jointly designed AI literacy programmes
• Ongoing measurement and evaluation

All of this, the company notes, happens while adapting to local languages, curricula and institutions. That qualifier matters. The report is explicit that country examples reflect their particular contexts and arrangements, and their findings should be interpreted within those limits.

The report organizes its lessons around five themes: clear educational goals, educator preparation, local partnerships, critical AI literacy, and credible evidence.

What are the three early patterns?

OpenAI identifies three patterns emerging from the partnerships so far. Each comes with a promise, and each, read closely, comes with a caveat.

First, country partnerships offer context-specific lessons. Experiences in Estonia, Greece and Slovakia point to three transferable principles: strong public leadership to set a clear direction, investment in teachers' skills and confidence, and practical support tailored to local curricula and classroom needs. Implementation will vary across countries, the authors concede, but the principles offer a foundation for governments seeking to integrate AI responsibly and at scale.

Second, the guidance gap is becoming increasingly urgent. The ThinkYoung figures — 79% of students using AI for homework, only 47% shown how — anchor this claim. Students are already using these tools. Whether they are using them well is a different question, and one the report admits remains largely unanswered.

Third, early benefits for teachers are clear — but only the early ones. Adoption and reported time savings for teachers are encouraging, the report says. But understanding where AI strengthens knowledge, independent thinking and retention requires sustained evaluation. The company says the government actions it observed point to how research in this area can be prioritized. Notably, "reported time savings" is self-reported survey data, not an independent outcome measure — a distinction the report's own caution about evidence invites readers to keep in mind.

The takeaway in plain words: teachers save time and students use AI heavily, but nobody yet knows whether learning actually improves.

How does OpenAI propose to measure progress?

The report acknowledges the evidence question directly. OpenAI says it is learning to judge progress through teacher adoption rates, reported time savings, surveys, programme monitoring and consultation — mechanisms that capture participants' experiences and the barriers they encounter.

That is monitoring, not proof. The report does not claim learning gains, and it says so: these are early experiences, "rather than a single model for education systems to follow."

To build the evidence base, OpenAI points to the OpenAI Learning Lab, an initiative it says brings education researchers together to understand when and how AI can support education systems. The company frames its mission — ensuring artificial general intelligence benefits everyone — as dependent on helping learners flourish, whatever their background or school.

"We have a responsibility to contribute to that evidence and learn openly from the findings, including when they challenge our assumptions," the report states. That sentence is doing real work: it is a commitment to publish inconvenient results, and the credibility of the entire enterprise will rest on whether that commitment holds.

What is not said?

Several things the report does not say deserve attention. It names no specific products, no pricing or procurement terms for the government partnerships, and no enrollment or usage figures for any country programme. The ThinkYoung survey's methodology — sample size, countries covered, dates — is not disclosed in the report itself.

The claim of "equitable access to frontier AI tools" is also left undefined. Whether that means donated access, subsidized licences or standard procurement is not specified. For governments weighing similar deals, those details are the difference between policy and press release.

The takeaway: the report is strongest on diagnosis and weakest on independently verifiable outcomes.

Why does this matter beyond education policy?

The context is a competitive market. AI companies are racing to embed their tools in public institutions, and education is among the most sensitive. A vendor simultaneously supplying the tools, funding the research and shaping the literacy programmes occupies an unusual position — one the report's own emphasis on "credible evidence" and "critical AI literacy" implicitly acknowledges.

The report closes on the stakes: "The choices made today will shape how the next generation learns, works and participates in society." With ambition, care, public leadership and a commitment to evidence, the authors argue, AI can help extend educational opportunity and strengthen young people's capacity to understand, create and act for themselves.

The next test is concrete: whether the sustained evaluation the report calls for produces published, independent evidence of learning gains — or whether the 79% of students already using AI will keep learning to use it on their own.