AI-focused private schools are expanding rapidly across the country, but a growing body of research suggests that artificial intelligence tutors are not actually superior to human instruction. Over the past decade, for-profit schools like Alpha School have grown from a single campus in Austin, Texas, to more than 15 locations nationwide, including in major cities like New York and San Francisco, with tuition ranging from $40,000 to $75,000 annually. Yet despite the hype and investment, the evidence tells a more complicated story about what AI can and cannot do in the classroom.

What Does the Research Actually Show About AI Tutoring?

The promise of AI in education rests on a compelling idea: personalized learning tailored to each student's pace and needs. Advocates argue that AI can provide every student with an individual tutor, something that has historically been available only to wealthy families. However, when researchers examine the actual outcomes, the picture becomes murkier. A major review published by the National Bureau of Economic Research in 2020 found that many different kinds of human tutoring produced consistent learning gains across subjects and age ranges. More recent studies from 2025 and 2026 have shown that AI tutoring can improve student achievement and that generative AI can modify tutoring sessions based on a student's progress. Yet extensive analysis of different studies found no significant difference in learning outcomes between computer tutors and human tutors when compared directly.

One widely cited 2025 study did find significant positive effects of AI tutoring on student achievement across subjects and grade levels. In another study, researchers at Harvard compared active learning in a physics class with a specially designed AI tutor built by the same professors teaching the course. Students using the AI tutor reported learning the material faster and feeling more motivated. However, these students were already highly motivated with good study skills, and the AI tutor was built specifically for their course by their own instructors. These conditions rarely exist in typical school settings.

How Are Schools Actually Using AI in the Classroom?

Alpha School and similar institutions replace traditional in-person instruction with personalized AI tutoring that condenses reading, math, and other subjects into a two-hour study period. This AI tutoring block is supplemented with in-person workshops and sessions led by "coaches" or "guides," who are not necessarily licensed teachers. These sessions can focus on nonacademic life skills like public speaking and entrepreneurship, as well as art and physical education. The model appeals to parents frustrated with what they see as a "one-size-fits-all" approach to traditional schooling, where all students study the same material but learn at different speeds.

Tech leaders have amplified this vision. Sal Khan, founder of Khan Academy, has argued that artificial intelligence could give every student an individual tutor responsive to their specific needs. Bill Gates has speculated that AI will replace many teachers within the next decade. However, education policy scholars caution that this framing creates a false choice between responsive, motivating AI programs and rote classroom lecturing, when the reality is far more nuanced.

What Makes Human Tutoring Effective That AI Cannot Replicate?

One-to-one tutoring with a human teacher, parent, or paraprofessional has been shown to be very effective in raising student test scores and improving overall learning. "High-impact" tutoring involving regular, long-term sessions in small groups held within schools has proven particularly effective. For tutoring to work, certain conditions must be met: tutors and students should meet consistently, tutors should be skilled and ensure their teaching integrates fully with the student's curriculum, and relationship-building between tutors and students is critical for keeping students motivated and focused.

Beyond academic metrics, young students need to develop strong social skills and what experts call "relational intelligence" to flourish in social settings. Human relationships are messy and filled with emotions. Parents misunderstand their children, kids get frustrated, families argue and reconnect. These imperfect interactions and the productive struggle they create are how children learn resilience, emotional regulation, flexibility, and how to navigate real relationships. Eye contact, shared laughter, patient answers to questions, and back-and-forth exchanges activate neural circuits designed for connection in ways that no algorithm can match.

How Can Schools Use AI to Support Rather Than Replace Teachers?

Rather than viewing AI as a replacement for human educators, education experts suggest a more promising strategy: using AI to support teachers in becoming better educators. A 2024 study showed student improvement in middle schools in low-income areas when human tutors had access to AI tutoring support. A 2026 study found that when teachers use AI for lesson planning, experienced teachers tend to critically revise the content to better link it to the overall curriculum. This suggests that the real value of AI lies not in automating instruction, but in augmenting teacher capabilities.

"Instead of trying to replace teachers with AI, I think that a more promising strategy is to support teachers in using AI to become better educators," explained Gerald K. LeTendre, Professor of Educational Administration at Penn State.

Gerald K. LeTendre, Professor of Educational Administration, Penn State University

The key insight is that teachers need support and professional development to make the most of these new and complicated tools. Rather than considering a future where AI tutors replace human teachers, education leaders should focus on what kind of training and resources can help educators integrate AI effectively into their practice.

What Are Teens Actually Saying About Learning With AI?

Young people themselves hold diverse views on AI in education. For some, AI serves as a helpful learning companion. Tessa Klein, an 18-year-old recent high school graduate from New Jersey, found AI helpful for providing feedback on essays and walking her through complex science concepts. She describes it as "an opportunity to have sort of like a private tutor that maybe other students cannot have or cannot afford." Charles Ansevin, a 15-year-old from Ohio, views ChatGPT as a friend with whom he has had "very meaningful, intelligent discussions".

She

However, other teenagers express serious concerns. Dorian Prado, 16, from Texas, says he is "very against AI," arguing that "it makes it to where thinking is optional, and that should never be the case. You don't think, you don't learn. It's making us dumber." For Dammie'on McColley, 18, from Indianapolis, the worries extend beyond education to broader economic impacts. He expressed concern that AI could "throw off jobs and things like that. That's people's only way of bringing in income to feed their families. And if we have machinery that's taking over that, then what are they going to do?".

What Are the Hidden Risks of AI Companions in Childhood Development?

Beyond the classroom, technology companies are developing AI companions that resemble intelligent toys and robots designed to interact socially with children. These systems can answer questions, help with homework, tell bedtime stories, and feel like friends. However, experts warn that AI companions may carry hidden developmental risks. Dana Suskind, a professor of surgery and pediatrics at the University of Chicago Medical Center, originally titled her book on AI and childhood "The Trojan Teddy Bear" to warn that AI companions may seem cute and cuddly but carry concealed risks for child development.

Suskind's research draws on neuroscience showing that young children's brain development is heavily influenced by back-and-forth interactions with parents and caregivers during the first years of life. Renowned developmental psychologist Patricia K. Kuhl proposed the "social gate" hypothesis, suggesting that children's brains are biologically primed to learn through social interaction. Studies show that babies learn language much better from a live person than from a screen because social interactions engage the brain in ways passive media does not.

Steps to Ensure AI Supports Rather Than Replaces Human Connection in Learning

• Prioritize Human Relationships: Ensure that AI tools supplement rather than replace direct interaction between students and qualified teachers or tutors who can build ongoing relationships and provide emotional support.
• Integrate AI With Curriculum: When schools use AI, experienced teachers should critically review and revise AI-generated content to ensure it connects meaningfully to the overall curriculum and learning objectives.
• Invest in Teacher Development: Provide educators with professional development and support to help them use AI effectively as a tool to enhance their teaching, rather than viewing AI as a replacement for their expertise.
• Monitor Social and Emotional Development: Be cautious about replacing human interaction with AI companions, particularly for young children, and ensure that kids maintain rich, messy, emotionally complex relationships with real people.
• Demand Transparency on AI Goals: Recognize that companies design AI systems with their own goals, which may include maximizing engagement, collecting data, and making money, not necessarily serving children's developmental needs.

The bottom line is that while AI can be a useful tool in education, it works best when it enhances human teaching rather than replaces it. Schools charging premium tuition for AI-centered instruction should be transparent about what research actually shows: AI tutors are not superior to human tutors, and the irreplaceable value of education lies in human connection, relationship-building, and the messy, productive struggle of real learning.

" }