AI in Education Surges as Experts Warn of Critical Thinking Decline and Learning Risks

AI Usage Among Students and Teachers Reaches Critical Mass

AI in education has reached a tipping point. An estimated 85% of K-12 public school teachers reported using AI during the 2024-2025 school year, primarily for curriculum and content development 1

1

. Among students, the numbers are equally striking. Nearly 80% of Australian university students now use AI in their studies, while a UK survey found 94% of undergraduates were using it to help with assessed work 4

4

. In the United States, 26% of teens reported using ChatGPT to complete schoolwork in 2025, up from just 13% in 2023 1

1

. An estimated 86% of K-12 students have used AI in general, with 50% reporting they use it for schoolwork such as learning about topics outside class, tutoring on specific subjects, or receiving help with homework assignments 1

1

. Among 12- to 17-year-olds specifically, 59% use AI to search for information and facts, according to Common Sense Media .

Weakening Critical Thinking Skills and the Performance Paradox

The rapid AI integration in classrooms has triggered alarm among education researchers who warn of serious risks to student learning. A phenomenon known as the "performance paradox" reveals a troubling pattern: while students' short-term performance on tasks may improve with AI assistance, their long-term learning suffers significantly 4

4

. A 2025 randomized experiment with high school students in Turkey demonstrated this effect clearly. Students using an AI assistant appeared to solve math problems more effectively in classroom tasks, but their actual learning collapsed as soon as the AI was removed during assessment 4

4

. This cognitive offloading from human to AI creates what researchers call an "illusion of competence," where students overestimate how much they have learned because the tool provides clear, polished responses that signal deep mental engagement is no longer necessary 4

4

. Almost two-thirds of parents of K-12 students said in 2025 that AI is weakening important academic skills their children need to learn, including writing, reading comprehension, and critical thinking 1

1

. The Brookings Institution warned in a 2026 report that risks of using generative AI in education overshadow its benefits, with students who offload thinking to AI doing less of it themselves, creating a compounding atrophy over time 5

5

. As one student told researchers, "It's easy. You don't need to use your brain" 5

5

.

Lack of AI Policies in Schools Creates Uncertainty

Despite widespread adoption, policies and teacher training have not kept pace with how frequently students and educators use AI. Only 35% of school district leaders reported in 2025 that they provided students with any AI training, according to the RAND Corporation 1

1

. Additionally, just 45% of principals reported having school or district policies or guidance on AI use in schools 1

1

. This lack of AI policies in schools leaves students navigating a vast gray area. Business students interviewed for a UK study knew that copying and pasting from ChatGPT would be considered cheating, but described widespread confusion about what constitutes legitimate assistance versus academic misconduct 3

3

. Different courses and lecturers gave different answers about whether students could ask ChatGPT for feedback on draft paragraphs or suggestions for alternative headings 3

3

. An average of 71% of K-12 teachers reported that when students use AI to complete schoolwork, it becomes difficult to determine whether student work is their own 1

1

. States are largely leaving districts to develop their own policies, districts are leaving teachers to figure it out, and many teachers received their first real AI training at Saturday workshops funded by companies whose products were being demonstrated 5

5

.

Challenges to Academic Integrity and Equity Concerns

Challenges to academic integrity extend beyond plagiarism detection. The shift raises fundamental questions about equity and access. Students who pay for premium versions of AI tools like ChatGPT feel they receive more accurate, detailed support than peers using free versions 3

3

. Some students view this as another form of educational inequality, where success on assessments increasingly depends on whether you can afford better algorithms and possess the necessary skills to prompt systems for optimal results 3

3

. The Brookings Institution noted this may be the first time in education technology history that schools will have to pay more for more accurate information, with free AI tools most accessible to under-resourced schools tending to be the least reliable 5

5

. However, some students argue AI can make education fairer. Students with dyslexia, ADHD, or other conditions described using ChatGPT to help with planning, time management, or turning rough notes into clearer sentences, while international students said it helped them write in more polished academic English 3

3

. Special education teachers are testing benefits too, with 57% reporting in 2025 that they use AI to help develop individualized education programs for students with learning disabilities 1

1

.

Long-Term Educational Impacts Demand Attention

The long-term educational impacts of AI remain deeply uncertain. Research from the Stanford Accelerator for Learning found that when students use AI and then are told they can no longer use it for studies, they actually perform worse than those who never used AI, demonstrating that additional research on how AI influences students' long-term learning and development is necessary 1

1

. Studies from 2019 through 2022 suggested AI might help student learning and motivation through personalized learning experiences, but evidence appears less promising when considering how students learn after they stop using AI 1

1

. The risks of AI in education extend to student safety as well. Recent examples include students who self-harmed or died by suicide after using AI for mental health support, and a 2025 study found chatbots responding to 60 simulated mental health scenarios sometimes made harmful proposals such as cutting off all human contact for a month or dropping out of school 1

1

. A study from Microsoft and Carnegie Mellon found that popular chatbots may actively diminish critical thinking skills, and AI systems designed to be agreeable turn out to be poor models for the friction that builds social and emotional resilience 5

5

. Seventy-one percent of parents and 60% of kids and teens believe that by the time young people are adults, people will be so dependent on AI that they won't be able to function without it . Global approaches vary widely. Estonia built a national AI literacy program that modified ChatGPT to respond to student queries with questions rather than answers, while Iceland is running a cautious pilot where teachers experiment with AI for lesson planning but students aren't involved at all 5

5

. In the United States, Microsoft and OpenAI have committed tens of millions of dollars to teacher training through the country's two largest teachers unions, and in Florida alone, more than 100,000 high schoolers now have access to Google's Gemini chatbot through their school districts 5

5

. Experts suggest universities and teachers must move from treating AI as an "answer oracle" to using it as a partner in thinking, offloading extraneous tasks like checking grammar while using AI as a "cognitive mirror" that asks clarifying questions to force students to engage in explanation 4

4

. Assessment design must evolve accordingly, and student privacy concerns remain unresolved as most deployments are outpacing research by a wide margin 1

1

5

5

.