79% of students use AI for assignments, but fewer than half received guidance

ThinkYoung surveyed 5,000 secondary-school students aged 13 to 19 and 600 teachers across 12 countries in Europe, the Middle East, and Africa. According to results cited by OpenAI, 79% of surveyed students use AI for homework or assignments, but only 47% said someone had shown them how.

Students moved 32 percentage points ahead of guidance

The gap between use and instruction shows that schools no longer decide whether AI enters education. They mainly decide whether students use it openly, critically, and under understandable rules. ThinkYoung's own page also says 95% of surveyed teenagers use some AI tool.

OpenAI connects the survey with partnerships in Estonia, Greece, Slovakia, and Italy. The programs emphasize tool access, teacher preparation, local curricula, and continued measurement. The company says these are early experiences rather than one template for every country.

Teachers need rules before another license

A school can buy ChatGPT faster than it can redesign assignments and assessment. The practical gap is therefore broader than learning how to write a prompt. Teachers need to know when AI supports independent thought, when it replaces student work, and how to ask learners to explain their process.

Local context also matters for ministries. Language, student age, curricula, and device access shape what can be introduced responsibly. A universal training package could easily confuse product use with AI literacy.

The tool vendor is also writing the usage guide

The figures come from ThinkYoung research supported by OpenAI, while the interpretation is published by a company selling AI access to schools. That does not invalidate the data, but it calls for caution. The public summary lacks country-level results and longitudinal evidence that use improves knowledge, independent thinking, or retention.

OpenAI acknowledges that uncertainty and calls for continued evaluation. Schools need to separate teacher time savings from student outcomes. The two may move in different directions.

Defensible work will reveal whether guidance helps

The next step must measure more than account counts and course completion. Schools need to test whether students can spot a model error, explain their choices, and solve a similar problem without AI. Only that evidence can show whether the 79-to-47 gap is closing through learning or merely through more product instructions.

Lilith's verdict

Students already sit at their desks with AI open beside the notebook. Schools now have to teach them how to think with it, not merely where to click.