The Leveller That Multiplies: How AI Tutoring Widens the Education Gap
There is a particular kind of promise that technology likes to make, and it goes like this: the thing that was once scarce and expensive will now be abundant and cheap, and so the people who never had it will finally get it, and the world will tilt a little closer to fair. It is a seductive story. It has been told about the printing press, about radio, about the personal computer, about the internet, and now it is being told, loudly and everywhere, about artificial intelligence in the classroom. A patient, infinitely available tutor for every child on earth. The end of the postcode lottery. The democratisation of quality education, finally, at scale. It is a beautiful idea. The trouble is that the evidence, as it accumulates through the early months of 2026, keeps pointing in the opposite direction. Not subtly, either. A run of peer-reviewed studies and institutional research published between December 2025 and February 2026 tells a remarkably consistent story, and it is not the story on the marketing deck. The story the research tells is that AI in education, deployed the way it is currently being deployed, is far more likely to widen the gap between advantaged and disadvantaged children than to close it. The leveller, on closer inspection, looks a lot like a multiplier. This matters enormously, and not only because education is the mechanism through which societies decide who gets to be what. It matters because the gap between the promise and the evidence has become a chasm, and into that chasm governments are pouring public money, vendors are pouring product, and children, millions of them, are being enrolled in an experiment whose results nobody has bothered to wait for. The question is no longer whether AI can help some students learn. It plainly can. The question is who it helps, who it leaves behind, and who is supposed to answer for the difference. The Study Nobody Wanted to Read at the Product Launch Begin with the most direct rebuttal to the…