No, AI Is Not Making Children 55 Percent Dumber

A headline doing the rounds this month claims that AI is making children 55 percent dumber. The figure has migrated from a single MIT Media Lab study into Substack threads, parenting newsletters, and the front pages of finance blogs. It is the kind of number that sounds precise enough to trust and frightening enough to share. Both qualities should prompt closer reading.

The 55 percent comes from Your Brain on ChatGPT, an MIT Media Lab paper that tracked EEG brain activity in university students while they wrote essays with and without ChatGPT. Participants who exclusively used the model showed roughly 55 percent less neural connectivity than those writing unassisted. The researchers also recorded weaker memory retention for the content of the essays and a diminished sense of authorship over the finished text.

What the number is, and is not

That measurement is not an IQ drop. It is not a claim that children become 55 percent less intelligent. It is not even a claim about children: the study sample was university-age. The figure describes a difference in real-time brain activity during a specific task, in a specific population, with a specific delivery method. The researchers themselves describe the effect as cognitive offloading, the same mechanism by which calculators reduce arithmetic effort and GPS navigation reduces mental mapping.

The leap from those findings to 24/7 Wall St.'s May 10 headline, "Generational Risk: AI Is Making Kids 55% Dumber," is large. Other coverage has been more careful. The Harvard Gazette framed the research as evidence of a tradeoff, not a doom signal, and noted that the same students recovered some cognitive engagement when they wrote a second essay without AI assistance.

The real finding

Strip away the framing and there is still a serious study underneath. The MIT team did find that heavy AI use in essay tasks reduces neural engagement, and that some of the effect persists even after the tool is removed. A March 2026 Harvard Business Review piece reported similar findings in workplace settings, with workers describing a fatigue they called "brain fry" after long stretches of AI-assisted work.

The honest reading is closer to this. Outsourcing thinking to a fluent model feels frictionless because most of the work happens elsewhere. The skill that produced the work also lives elsewhere, and skills that are not used erode. Students who rely on AI to draft every essay will, over time, get worse at drafting essays.

That is a meaningful concern, especially given the March 2026 RAND figure that 62 percent of middle school, high school, and college students use AI for homework, up from 48 percent seven months earlier. It is not, however, the same concern as the headline. "Using a tool can let your skills decay" has been true since the abacus. "AI makes children 55 percent dumber" is a different and louder claim, and the evidence does not support it.