Kids, education, and cognitive skills

🔍The Reality

The largest measurement yet arrived on September 8, 2026, and it cuts against the optimistic reading above — though not in the direction most coverage suggested. The OECD published PISA 2025, the first round run since AI use became ordinary in schools, covering more than 760,000 fifteen-year-olds who represent about 33 million students across 91 countries. After adjusting for socioeconomic status, students who never use AI generally outperformed those who do in science. But the shape of the finding matters more than the headline. What students used it *for* separated the results: summarising a text they were assigned to read, or drafting a writing assignment, tracked worst, while using it for preliminary research or general help learning tracked far better. Frequency was a curve rather than a line, with yearly and daily users faring worst and monthly or weekly users best. And students who had been taught to judge the quality of what an AI tells them outperformed non-users outright — which is the one result that vindicates the "thoughtfully" qualifier in the myth section rather than undercutting it. The OECD's own framing is that learning is a productive struggle: where technology enables that struggle students advance, and where it short-circuits it they do not. Reading and mathematics scores across the OECD have fallen sharply over the past decade, a longer trend than AI and not attributable to it. Unsupervised AI use by young students can short-circuit the productive struggle that builds skill. Schools that haven't updated assessment design face genuine cheating problems — and the scale is no longer theoretical. In mid-2026 Mexico's largest university ran its entrance exam fully online with AI proctoring; about 16 percent of candidates scored 100 or more out of 120, against roughly 3 percent in the five prior years, and about 58,000 applicants had to retake it in person. Excessive screen time and chatbot-mediated relationships raise real questions for child development. Districts are now acting on that concern rather than debating it: on September 2, 2026 New York City — the largest school system in the United States — imposed a one-year moratorium on classroom AI use through eighth grade, covering roughly 600,000 students, banning companion chatbots in every grade and barring teachers from using AI to grade work. Notably, it is a pause rather than a prohibition, and it is not uniform: high schoolers keep limited access plus AI literacy classes twice a year, teachers may still use AI for lesson planning, and exceptions cover students with disabilities, multilingual learners and computer-science courses. Regulators are now moving on the platform side as well as the classroom side. On September 8, 2026 Australia released draft Digital Duty of Care legislation extending its social-media age ban into a general safety obligation: online games, apps and AI chatbots would have to shield users under 18 from addictive design features and from content promoting eating disorders, misogyny or self-harm, with the eSafety Commissioner empowered to order removal of nudify apps and penalties reaching 109.2 million Australian dollars. It is a draft for consultation rather than passed law, but it is the first national framework to place AI chatbots inside a general online duty of care rather than regulating them separately as AI. Parents and educators are right to want clearer guidance, and the guidance is starting to arrive.