AI in Schools Creates a Two-Tier System as Reading Scores Hit Record Lows

WASHINGTON — A deepening crisis in basic literacy and cognitive development is colliding with the rapid, unregulated integration of artificial intelligence in American classrooms, sparking a polarizing divide between under-resourced public school districts and wealthy families seeking high-tech private workarounds.

The immediate academic boost provided by AI tools often masks a long-term erosion of actual learning, according to recent empirical research. A study tracking AI adoption among 26,811 middle and high school students in grades seven through 12 in China found that while utilizing AI tools boosted homework scores by 18%, it subsequently triggered a 20% drop in unassisted exam scores.

At the Alpha School, a private institution expanding its reach across the country, tuition reaches up to $75,000 per year. The school, which currently serves more than 1,200 students, has announced plans to scale its operations to 50 campuses this year. Under this model, traditional teachers are replaced by human “Guides.” These Guides do not design lesson plans, deliver lectures, or grade homework; instead, their primary responsibility is to monitor student motivation and provide emotional and behavioral support.

Despite these eroding foundational skills, generative AI adoption among adolescents has skyrocketed. Data from 2025 surveys conducted by the College Board, the nonprofit organization that administers the SAT and Advanced Placement exams, reveals that an estimated 84% of high school students now use AI tools for homework assistance. Meanwhile, institutional oversight remains lagging; one in five high schools permits the use of AI but has failed to establish any formal policy governing its application.

The educational landscape is shifting at a time when student achievement has hit historic lows. According to the latest results from the Programme for International Student Assessment (PISA)—a globally recognized benchmark administered by the Organisation for Economic Co-operation and Development (OECD)—reading and math scores for 15-year-olds across 38 free-market democracies reached their lowest average ever recorded. In the United States, average reading scores dropped by 14 points to 490, marking the lowest performance level since the country began participating in the PISA exams in 2000.

This disconnect highlights a growing concern among neuroscientists and cognitive researchers, who warn that outsourcing intellectual labor to digital assistants short-circuits the mental processing required to retain information. Testifying before Congress on the impact of technology in education, neuroscientist Jared Cooney Horvath warned that screen-based learning has consistently distracted students, resulting in lower test scores. “Our kids are less cognitively capable than we were at their age,” Horvath told lawmakers, emphasizing that the widespread introduction of computers and tablets in classrooms—a movement that accelerated rapidly after 2014—has yielded counterproductive results.

The policy response from public school administrators has been highly fragmented. Fearing cheating and cognitive dependency, some of the nation’s largest public school systems have enacted restrictive measures. New York City Public Schools, the largest district in the United States, recently announced a one-year ban on student-facing generative AI tools for children in programs ranging from 3-K through the eighth grade.

Huriya Jabbar, a professor of education policy at the University of Southern California, points out that high-poverty public school districts face structural barriers that prevent them from safely and effectively integrating AI. Public schools frequently suffer from high teacher turnover rates, making continuous professional development on emerging technologies difficult to sustain. Additionally, these schools face severe pressure to raise standardized test scores, leaving little administrative bandwidth or budget for experimental teaching models. “AI—just like any other tool, technology, curriculum—is going to play out in an already unequal landscape,” Jabbar said.

“This is not a debate about rejecting technology,” Horvath noted in his prepared remarks. “It is a question of aligning educational tools with how human learning actually works. Evidence indicates that indiscriminate digital expansion has weakened learning environments rather than strengthened them.”

For wealthy parents, this approach is viewed as essential preparation for an economy likely to require high AI literacy and feature fewer entry-level hiring opportunities. “I definitely think using AI well is a skill in the future, if not the skill, so it’s kind of a disservice not to have that in school,” said Sarah Cone, a venture capitalist whose 8-year-old daughter attends the Alpha School campus in New York.

The Alpha School’s model represents a radical departure from traditional pedagogy. Students spend two hours each morning working directly with personalized AI tutors on core academic subjects. The afternoon is reserved for hands-on, collaborative workshops designed to build interpersonal skills, leadership, and public speaking.

This decline in fundamental capabilities has followed students into higher education. University professors report that incoming high school graduates increasingly struggle to analyze long-form texts, frequently relying on generative AI to summarize reading assignments. The decline in deep reading is reflected in generational metrics: in 2025, Americans between the ages of 18 and 25 reported reading an average of just 5.8 books per year, compared to 8.2 books for those aged 30 to 44, and 12.1 books for seniors aged 65 and older.

The ability to curate a balanced, highly supervised AI curriculum remains a luxury of the wealthy, leaving public school systems to navigate the technological shift with far fewer resources.

While researchers agree that indiscriminate AI use can damage cognitive development, some educational experts suggest that intentionally designed, highly specific AI tools—such as interactive chatbots designed to act as Socratic math tutors rather than simply supplying correct answers—can support student learning. However, implementing those sophisticated guardrails requires a level of technological infrastructure and pedagogical oversight that many public school districts currently lack.

As the division between high-tech private instruction and restrictive public policies widens, educational advocates emphasize that the primary goal of schooling must remain focused on developing critical thinking. “We want readers who can interpret, question, make connections, recognize perspective, evaluate evidence, sit with ambiguity, and form their own understanding,” said Kirsten Peterson, a senior project manager at the nonprofit Education Development Center. “We have to be really careful not to use AI to remove the very thinking students need to practice in order to become literate.”