The AI-driven school network charging parents up to $75,000 a year is falling short of its central promise that students learn twice as fast, even as it pushes into public education.
Only 42% of Alpha School students met the network's own goal of learning twice as quickly in both math and reading. Math is the wider gap: about half of K-8 students at its private campuses missed the math target, while roughly one in five missed their reading goals. Students who started behind fared worst, with just 37% of those scoring at or below the 50th percentile nationally in math reaching the 2x growth mark by winter.
Alpha has no traditional teachers. It staffs classrooms with 'guides' who neither plan lessons nor grade work, while students spend two hours a day on TimeBack, an AI tutor that cofounder MacKenzie Price says personalizes and accelerates learning. The rest of the day is reserved for workshops on skills like public speaking and relationship building.
The pitch has won influential backers, from Silicon Valley figures like Reid Hoffman and Bill Ackman to U.S. Education Secretary Linda McMahon, who toured Alpha's Austin campus in September 2025 and called its model the most exciting thing she had seen in education in a long time.
Now the network is testing that model on far tougher ground. More than 2,600 public school students in Texas and South Carolina use Alpha's platform, and Houston's district is piloting the software at two campuses. Public schools cannot shed struggling students and lack the prize incentives Alpha leans on at its private sites, making the results a harder measure of what the model actually delivers.
The two-hour premise is already under strain. Guides and parents say students work late into the night to hit the school's goals, with one describing the pressure as stressful and anxiety-inducing for a 13-year-old. Former employees also allege the AI produced faulty lesson plans and that the company scraped other online courses without permission.
Broader research deepens the doubts. A study of more than 26,000 secondary students in China found homework scores rose 18% after adopting generative AI while closed-book exam scores fell 20%, and the OECD's latest PISA results found daily AI users scored lower in science than nonusers. University professors warn that students leaning on AI summaries struggle with sustained reading, and in some cases with reading full sentences at all.