Students Know AI Is Hurting Their Thinking. Schools Keep Expanding AI Programs.

Last updated: July 5, 2026

Your child is probably already using AI for schoolwork. According to a March 2026 RAND study of more than 1,200 students, 62 percent of middle and high schoolers now use AI to help with homework, up from 48 percent just ten months earlier. What those same students said next is the part most school AI programs are not addressing: 67 percent believe that the more students use AI for schoolwork, the more it harms their critical thinking skills. That number climbed more than 10 percentage points in a single school year. The students are watching it happen in real time. Meanwhile, schools are racing to add AI programs, and nearly all of those programs teach children how to use AI rather than how to understand it. That difference is not a minor curriculum quibble. For children still building the cognitive foundation that learning depends on, it is the whole question.

TL;DR

RAND Corporation (March 2026, 1,200-plus students, Gates-funded): 62 percent of middle and high schoolers use AI for homework, up from 48 percent in May 2025.
67 percent of those same students say AI harms critical thinking, up 10-plus percentage points in a single school year; 75 percent of female students agreed.
A 2026 ScienceDirect analysis found a significant negative correlation between AI tool usage and critical-thinking scores, with younger students showing higher dependence and lower scores.
Dharav High School (India) and the CBSE 2026-27 national mandate treat AI literacy as cognitive skill development: pattern recognition, bias detection, algorithmic reasoning.
For children still building reading and processing skills, AI tool dependency risks bypassing the exact cognitive practice those systems require to develop.

New RAND data finds 67 percent of students say AI is eroding their critical thinking skills. Here is what parents need to ask about their child’s school’s AI program.

Common questions

What is the difference between AI tool use and real AI literacy?

AI tool use teaches students how to generate outputs from AI: how to prompt it, which apps to use, and how to evaluate whether the output is useful. Real AI literacy adds a layer underneath that: understanding how AI systems make decisions, what pattern recognition means, where training data comes from, how bias enters outputs, and when to question rather than trust a result. Schools teaching only tool use are creating users. Schools teaching how AI works are building the critical evaluation skills that make using AI something other than dependency.

My child uses AI to help with reading or writing. Is that harmful?

It depends entirely on the use. A child with reading challenges using text-to-speech to access content while still receiving daily structured reading instruction is being scaffolded appropriately. A child using AI to decode words, rephrase sentences, or answer comprehension questions in place of doing that cognitive work is bypassing the practice those skills require to develop. The useful question is: is the AI freeing your child to think harder, or thinking for them on the skills their brain still needs to build? A screener is a starting point, not a diagnosis. If your child might need formal accommodations such as an IEP or 504 plan, or you suspect a vision, hearing, or medical cause, a professional evaluation is the route to those supports.

What should real AI literacy look like for younger students?

India’s national CBSE curriculum mandate for 2026-27 gives a concrete model: mandatory AI and computational thinking from grade 3, with focus on logical reasoning, pattern recognition, algorithmic thinking, and bias detection. Dharav High School’s approach centers on teaching students how AI systems work, including where their decisions come from and what questions to ask of their outputs. For younger students, age-appropriate versions of these concepts, recognizing patterns, questioning why a system produced a result, noticing when an answer seems off, build the cognitive foundation that genuine AI literacy requires.

What should I ask my child’s school about its AI program?

Three questions worth asking: First, does the program teach how AI works, including pattern recognition, training data, and algorithmic bias, or does it focus on tool use and prompt engineering? Second, for students who are still developing core processing skills such as reading, attention, and working memory, is there an explicit policy on which AI applications are off-limits during practice tasks? Third, what is the school’s written protocol for when independent cognitive work is protected from AI assistance? RAND researchers found schools need to be proactive and deliberate; the answer “we are still working on it” is worth following up with a specific timeline.

67% of students say AI is hurting their critical thinking. Up 10 points in one school year. Meanwhile schools keep adding AI programs that teach tool use, not AI thinking. RAND 2026. Tweet Now Follow Learning Success

What the data shows, and what one school is doing differently

The RAND findings from March 2026, funded by the Gates Foundation, are direct: students need schools to be explicit about when and how AI is used, with guidance that helps them use AI in ways that deepen rather than erode critical thinking. 67 percent of students already endorse the harm. 75 percent of female students said AI harmed critical thinking, compared with 59 percent of male students. The self-report lines up with research: a 2026 ScienceDirect analysis found a significant negative correlation between AI tool usage and critical-thinking scores, with younger students showing both higher AI dependence and lower scores.

Against that backdrop, a school in India is running a different model. Dharav High School, whose chairperson Devyani Jaipuria has publicly argued that over-reliance on technology in early education “can limit imagination and hinder the development of essential social and emotional skills,” focuses early AI education on how AI systems actually work: pattern recognition, algorithmic bias, and where automated decisions come from and what questions to ask of them. Students learn to evaluate AI outputs, not only generate them. India’s CBSE is scaling that framing nationally: starting in the 2026-27 academic year, AI and computational thinking become mandatory from grade 3, with curriculum centered on logical reasoning, problem-solving, pattern recognition, and bias detection. That curriculum treats AI literacy as a cognitive skill to be built, not a tool to be adopted.

Author Quote "

Digital tools have significantly enriched the learning experience. But over-reliance on technology, especially in early education, can limit imagination and hinder the development of essential social and emotional skills.

What the coverage gets wrong

Most education media frames the AI-in-schools story as schools trying to catch up to student use and teach responsible tool adoption. The RAND data complicates that frame: students are already using AI at scale, they already feel it eroding their thinking, and the gap is between tool-use programs and genuine AI literacy. Pattern recognition, critical evaluation of algorithmic outputs, and understanding of how AI systems make decisions are cognitive skills, not settings to configure. India’s national curriculum mandate treats them that way; most U.S. school programs do not. The practical difference for a parent: ask not whether the school has an AI program, but what cognitive skills the program builds rather than outsources.

Schools have made this mistake before, with a different label

There is a pattern in how school systems adopt new things. Learning styles, the idea that teaching to a child’s preferred mode improves learning, were tested and found without credible support in 2008, independently reconfirmed in 2024. A 2020 review of educators across 18 countries found nearly 9 in 10 still teach to them. Three-cueing, which trains children to guess words from pictures, context, and meaning rather than letter-sound relationships, was identified by cognitive scientist Mark Seidenberg as descriptive of how struggling readers read, not strong ones. It took a national investigative podcast in 2019 to begin changing laws, and most of those laws are only a year or two old. The system problem is consistent: schools adopt what is visible and comfortable, often long before the evidence has resolved.

Most school AI programs right now are tool-use programs. They teach students to generate with AI, prompt it, and recognize useful outputs. That is a real skill. It is not the same as understanding how a pattern-recognition algorithm makes decisions, where its training data originates, and when to trust or push back on its outputs. The former creates users. The latter creates thinkers. For children whose foundational cognitive systems are still developing, pattern recognition, working memory, processing speed, and phonological processing, the distinction matters more rather than less. Reading is not instinctual the way speaking is; it is a roughly 5,000-year-old invention the brain must be explicitly, systematically taught to perform (Stanislas Dehaene’s neuronal-recycling framework; David Geary’s biologically primary versus secondary knowledge). The International Dyslexia Association’s 2025 definition describes reading as multi-system, drawing simultaneously on language processing, phonological awareness, working memory, attention, and processing speed. Those are the same cognitive systems that AI tool dependency is found to erode. An AI program that bypasses pattern recognition in service of producing the right answer is running the wrong direction for children who most need to build it. Cognitive micro-skills are not a philosophy. They are the architecture reading and thinking are built on.

Key Takeaways:

Students already sense it: 67 percent of middle and high school students told RAND in March 2026 that AI use harms critical thinking, up 10-plus points in one school year.

Tool use is not the same as AI literacy: Dharav High School and India’s CBSE mandate focus on pattern recognition, bias detection, and algorithmic reasoning as the cognitive skills genuine AI literacy requires.

Ask the specific question: The useful question for your school is not whether it has an AI program, but whether that program builds cognitive skills or outsources them.

What this means for your child, and what to ask the school

This is not an argument against AI in schools. A child with dyslexia using text-to-speech to access a history chapter, while still receiving daily structured reading instruction, is being scaffolded, not harmed. The useful distinction is whether AI is freeing your child to think harder about something, or doing the cognitive work your child most needs to practice. RAND researchers put the burden on schools to make that distinction explicit. Most schools have not yet done that work.

Three questions worth bringing to your child’s school: First, does the AI program cover how AI actually works, including pattern recognition, training data, and algorithmic bias, or does it focus on tool use and prompt engineering? Second, for students still developing core processing skills, is there a policy on which AI applications are off-limits during their practice tasks? Third, what is the school’s written protocol for when AI is permitted versus when independent cognitive work is protected? The 67 percent of students who told RAND that AI harms their thinking did not need a study to reach that conclusion. The question is whether schools are moving as fast and as deliberately as the students already have.

Author Quote "

These findings suggest schools need to be explicit about when and how AI can be used.

The cognitive skills your child is building now, pattern recognition, critical evaluation, working memory, the ability to ask whether a source or an answer holds up, are not AI skills. They are thinking skills that AI use either strengthens or quietly borrows. The system that teaches tool use in place of those skills is not preparing your child for an AI-integrated future; it is preparing them to be dependent on one. That is a real problem, and 67 percent of students are already naming it. If you want to know exactly which cognitive systems your child is building and have a plan to strengthen each one before a tool fills the gap, the All Access program gives you that roadmap, with a free trial and a personalized Action Plan you keep regardless.