By 2035, most secondary students in many countries will have an AI study helper, and schools will weight in-class work more

Prediction: by 2035, in a large number of countries, most secondary-school students will have routine access to an AI study helper, and many schools will respond by giving more weight to work completed in class and to spoken and practical assessment, less to unsupervised homework.

Reasoning: tools that explain, quiz and give feedback are cheap and improving. The old finding that one-to-one tutoring is far more effective than a standard class, associated with Benjamin Bloom, creates a strong appetite for anything that approximates it. At the same time, if homework can be done by a tool, its value as evidence of a student's own understanding falls, which pushes assessment towards the classroom.

Confidence: low to moderate, about 50 percent. Access will be unequal, and the quality and effects of such tools are still being studied.

What would prove us wrong: if routine access remains rare or if schools do not change assessment practices.

How to check: national surveys of student technology use and assessment policies in 2026 and 2035.