Every conversation about artificial intelligence in schools asks the same question: How do we keep up? Districts buy tools, write acceptable-use policies, and train teachers on prompts. The race is to ensure students can use the technology. Almost no one asks what matters more: What is it doing to the minds we are educating?

The early evidence is not reassuring. At the MIT Media Lab, researchers used EEG to monitor brain activity while students wrote. Students who composed essays with ChatGPT showed measurably weaker neural connectivity than those who wrote on their own, and most could not quote a line from work they had finished minutes earlier. The researchers called the pattern “cognitive debt,” the bill that comes due when a machine does the thinking.

A Wharton field experiment with nearly 1,000 high school math students found the same trap. With AI tutors, students sailed through practice. When the tools were removed, those who had leaned on them scored 17 percent worse than peers who never used them. Most striking: They believed AI had helped them learn.

AI does not feel like it is hurting learning; it feels like progress. The erosion remains invisible until the tool is gone and the skill is gone.

These findings add to an attention crisis we have not solved. Psychologist Gloria Mark, who has studied human focus for two decades, found that the average time a person spends on a single screen before switching fell from 2.5 minutes in 2004 to 47 seconds in 2023. It is likely worse today. We ask children to do the slow, effortful work of learning in an environment built to interrupt them every 47 seconds. AI, with its instant answers, pushes in the same direction.

So what is education’s real job now? It is not to keep students away from AI; that ship has sailed, and these tools will shape their working lives. It is to ensure that when a student picks up the most powerful cognitive shortcut ever invented, a strong mind is at the controls, capable of sustained attention, working memory, reasoning, and discernment to recognize when the machine is wrong.

That means treating human cognition as core infrastructure, not a byproduct of covering content. The capacities AI erodes, such as focus, memory, metacognition, and tolerance for productive struggle, are trainable. One of the most direct ways to train them is daily mindfulness practice: Decades of neuroscience show that brief, consistent mindfulness strengthens attention and self-regulation and increases gray-matter density in regions tied to learning and memory. It is not a wellness trend but exercise for the exact mental muscles AI invites students to stop using.

The case is strongest in the early years. Attention, impulse control, and the capacity to sit with difficulty are laid down in childhood and adolescence, when habits of mind take hold for life. A few minutes a day of training in focus and self-regulation is not time taken from learning; it is what makes learning possible, and the earlier it starts, the deeper it takes root.

This is not theoretical. Thousands of schools have already built a few minutes of mindfulness practice into the day, no new subject required, and watched attention and self-regulation improve. But it remains the exception. At the very moment AI pulls hardest on young minds. The practice proven to strengthen them reaches less than 10% of students and is missing for the rest and belongs in every school, not a fortunate few.

The people building AI seem to grasp the point. Sam Altman, who runs OpenAI, says intelligence and awareness seem to go together. If the architects of these systems believe awareness must scale alongside machine intelligence, educators cannot treat it as optional.

Consider the irony. While schools turn to AI to support student mental health, AI researchers and developers have turned to mindfulness to steady the machines: A 2025 study found that calming, meditative prompts reduced ChatGPT’s anxious, biased responses to disturbing inputs. We are teaching the bots to regulate; we might do the same for the children using them.

We can keep optimizing for output and raise a generation fluent with AI but unable to think without it. Or we can build the cognitive muscle to use it well, beginning with a few minutes of daily mindfulness in every classroom, from the earliest grades.

The next race in education is not about better algorithms. It is about the minds that direct them. Districts that train attention and self-regulation as deliberately as they teach reading and math will graduate the students who lead in an AI world, not the ones it leaves behind.

(In the interest of full disclosure, this is work my company does.)