Universities abandon AI detection as research reveals technology boosts homework but crashes exams

Universities drop AI detection tools over fears about accuracy

When Orion Newby, an Adelphi University student, was wrongly accused of violating academic integrity by using AI to generate an essay, he faced a long battle to clear his name. In January this year, a New York court ruled in his favour. The judgment recorded that while the Turnitin AI detection tool used by the university had delivered an "AI-generated score of 100 per cent", Newby had submitted evidence from other AI detection tools purporting to show a zero per cent chance of AI-generated content, but the university upheld the misconduct finding. The court found that the university had failed to follow its own disciplinary procedures and had denied Newby a meaningful appeal. The Turnitin result is a probability score indicating that the tool concluded the submission was highly likely to have been AI-generated. But the ruling underscores concerns over the reliability of AI detection tools being used by universities and raises questions over the whole system of assessment. Institutions now face a dilemma following a surge in clandestine AI use by students. A recent study by researchers at Edinburgh Napier University, based on a survey of more than 6,600 students across seven UK universities, found that 32 per cent admitted some level of unpermitted AI use in assessments. Universities, determined to maintain integrity following the release of OpenAI's ChatGPT in 2022, have turned to applications with AI-detection features such as GPTZero, Copyleaks and Turnitin. The tools analyse features such as text structure and rhythm to identify non-human patterns, but their reliability -- specifically regarding false positives and bias -- is under fire. Several institutions have since restricted or disabled their use, including Vanderbilt, Yale, Johns Hopkins, Northwestern University, the University of Waterloo (Canada), the University of Cape Town (South Africa), and Curtin University in Australia. Edward Watson, vice-president for digital innovation at the American Association of Colleges and Universities, notes that the primary concern is the potential for false positives. "AI detection should, at most, serve a minor role in academic integrity cases," he says. "Faculty [should] never use AI detection as 'hard evidence' or 'smoking gun proof'." Annie Chechitelli, chief product officer at Turnitin, maintains that its detector is a "starting point" and "data point" rather than definitive evidence. She says the company continues to engage with universities to address concerns about "overuse or misuse", and last year launched a product allowing teachers to observe the evolution of a student's writing process, rather than relying solely on analysis of the final submission. Across the sector, reliability concerns are forcing a shift towards different kinds of assessment. However, Sam Illingworth, a professor of critical AI literacy at Edinburgh Napier University, notes that while some universities are redesigning assessments around oral components and student process, many still rely on detection. "AI detection tools are not the solution," says Judy Williams, pro vice-chancellor for education and students at Queen's University Belfast. "The technology is still developing, false positives can be unacceptably high, and AI-generated text can easily be modified, making detection unreliable. "If we want confidence in academic integrity, the answer is good assessment design," Williams says. "The important question isn't: 'How do we stop students using AI?' It's: 'What are we actually trying to assess?'" Despite this shift in attitude at Queen's and other universities, policies remain inconsistent. Illingworth's research auditing 163 UK universities found that more than 40 per cent had no publicly accessible AI policy. The lack of a clear approach fuels student anxiety; a December 2025 study by the UK's Higher Education Policy Institute found that 42 per cent of the 1,054 students who participated said they were less likely to use AI because they feared being falsely accused of cheating. The UK's Office of the Independent Adjudicator for Higher Education reports that while complaints remain low, AI-related misconduct cases are rising within internal university procedures. Chief of staff Adam Waddingham says: "Detection tools may be part of the evidence considered, but they should not normally be treated as determinative on their own." Urszula Lis, an executive committee member of the European Students' Union, worries that systems operate in a "closed" way, in which students cannot see how an AI tool reached its conclusion. "The bigger question is whether higher education should rely on detection-based approaches at all," she says. "We don't want academic integrity to become a surveillance exercise where the objective is simply to catch students doing something wrong." Two students studying in different countries may use AI in exactly the same way, yet one could be punished while the other would not Other criticisms include that the detection tools disproportionately deliver false positive results to non-native English speakers, as a widely cited 2023 Stanford study shows. Joachim Steinberg, a lawyer at law firm Crowell & Moring, specialising in tech litigation, says: "The greatest challenge is keeping institutional guidance and disciplinary processes aligned with the pace of technological change. While AI capabilities continue to evolve rapidly, universities are still working to establish clear, consistent and fair policies, leaving many students and educators operating in an environment of uncertainty." Globally, guidance remains fragmented. "It is much more dependent on the institution, which is frustrating because there is no consensus for effective practice," Illingworth observes. Lis adds that in Europe, the lack of national guidelines creates equity issues: "Two students studying in different countries may use AI in exactly the same way, yet one could be punished while the other would not." Illingworth's report concludes that many university policies function as compliance instruments: "They promise critical thinking but deliver audit trails. They name support yet deliver surveillance." "The durable answer is to design assessment around the thinking we want to see," he says. As AI use grows, the challenge for universities may be to move beyond surveillance to define what truly constitutes acceptable use. Nirmal Thacker, chief executive of GradPilot, an AI tool that helps prospective students with university applications, agrees: "I would like for universities to rethink assessment, welcome disclosure and be open about policy."

Universities face difficult choices over how to integrate AI

Universities around the world are facing a dilemma over the use of AI: whether to embrace the technology and train students how to use it, or tread cautiously with what is often an expensive and untested innovation. The approach chosen will shape the fortunes of the institutions and their students in the coming years. "There is a need to be purpose-driven. What is AI going to deliver to our students, to our staff? There needs to be strategic thinking around governance, ethics, and use cases," says Rose Luckin, a consultant on AI in education. Some are embracing AI. Ohio State University, for example, has pledged to embed AI across its curricula, ensuring that its undergraduate class of 2029 will be fluent in the use of AI when they graduate. But for others, the growing costs of AI use are an increasingly important drawback. "If you're being faced with the need to do greater staff capability training, greater student capability training and you've got a funding crisis. How do you equate those?" Luckin says. Those concerns about financial implications mean some universities are trimming their AI ambitions. Kemas Muslim Lhaksmana, dean of the school of computing at Indonesia's Telkom University, says the country could fall behind its South-east Asian neighbours in the adoption of AI. He hopes in the future there will be "light" versions of large language models and generative AI that can be run locally and will be cheaper to access for global south countries. Similarly, Reynald Cacho, professor of educational management at Philippine Normal University, says many students attending Philippine state universities are using the free versions of AI, limiting their ability to make the most of the tools. Divisions are also being seen in Sub-Saharan Africa. Sioux McKenna, professor of higher education studies at South Africa's Rhodes University, who also works in Rwanda and Kenya, says: "Unless you're in Nairobi, you're not going to have the bandwidth. While in South Africa the universities generally have continuous online access, the digital divide is getting bigger," says McKenna. Some universities are exploring possibilities in particular subject areas. Chie Adachi, whose remit is AI for education at the UK's Queen Mary University of London, has recently overseen a year-long initiative to integrate AI literacy into the curriculum framework of the medicine and dentistry undergraduate programmes. The programme aimed to ensure students had a foundational understanding of AI, and covered clinical applications and the use of AI in the workplace. "It can be a combination of conversations about what generative AI is, through to what they can do with AI for modelling diseases, the bioinformatics, or computing predictive models for certain cancers," says Adachi. Educators have to accommodate different attitudes towards AI and varying skill levels. Some students arrive at university knowing more about using AI than their lecturers. In contrast, there are others who have never encountered the tools before, and those who choose not to use them, citing concerns around inherent bias in the models, data security, hallucinations, the potential environmental impact and fear of being accused of cheating. A study carried out this year by Queen Mary Students' Union found around 20 per cent of respondents did not use AI at all, citing worries around ethics, misinformation and accusations of academic misconduct. In cases where students refuse to use AI, Queen Mary's staff are encouraged to find alternatives, such as small language models, which are locally hosted and have a lower environmental footprint compared with LLMs, Adachi says. This can help to alleviate some student concerns. Yet for many first-year students entering lecture halls for the first time, AI is already part of their lives. OpenAI's ChatGPT has been an accessible tool for their school studies for years. A 2025 Pew Research study of US teens aged between 13 and 17 years old found 64 per cent had used AI chatbots in their daily life. Students are also integrating AI into their studies. A survey by Queen Mary published this year found 43 per cent of its students were using AI at least weekly for academic purposes, and a further 25 per cent used it daily. More targeted aims, such as the teaching of vibe coding, which uses natural language interactions with AI to build software, are now a live topic of discussion at some universities. Kemas says vibe coding is not taught at Telkom University, because he thinks students can learn it themselves. Instead, the university teaches the fundamentals of real coding in the first year so they later understand when and how to use vibe coding. Similarly, while welcoming the use of AI, McKenna believes universities need to remember they are not technical schools, and their role is not to train students to vibe code or prompt in preparation for the world of work. As evidence rises of the increasing diversity in approaches to teaching AI at university level, it is unsurprising that regulators are beginning to respond. For example, in Australia, the Tertiary Education Quality and Standards Agency, which regulates higher education, has issued guidance based on surveys of institutions suggesting, among other recommendations, that staff are supported and trained in AI. Singapore too has established a Committee for Artificial Intelligence in Higher Education to create a system-level approach for institutions to identify issues together as AI evolves. Luckin says she has worked with Australia's higher education network HEDx to bring together groups of vice-chancellors including on the subject of AI. "We can work together to think through how to get the best results for students," she adds.

Artificial intelligence is exposing what's wrong with modern education

Artificial intelligence is exposing weaknesses in modern education that existed long before chatbots arrived. A new paper argues that many of the skills students are still graded on, including summarizing information and writing basic essays, can now be performed by AI tools in seconds. Rethinking what students should learn Professor Yong Zhao at the University of Kansas (KU) has spent years asking why schools barely change no matter how hard people try. He argues that the real problem is not students using artificial intelligence, but schools continuing to assess work that technology can easily replicate. If AI can successfully complete the assignment, it may be the assignment that needs to change. This has sparked debate over bans, cheating policies, and AI detectors. The paper suggests they may be overlooking a larger opportunity to rethink what students should learn and how they demonstrate it. AI exposes old problems Professor Zhao's new paper opens with a blunt claim. The real trouble with AI in classrooms, he writes, is not the technology. Generative AI can now do much of what schoolwork asks students to do. The tools summarize readings and turn a plain prompt into a passable essay in seconds. When the assigned task is one a chatbot can finish, students reaching for the chatbot stops looking like a character flaw. Outdated goals remain A recent survey found that nearly six in ten U.S. teens think cheating with AI happens regularly at their school. For Zhao, the pattern reveals a deeper problem than cheating. The goal itself has gone stale, because schools still reward the kind of work a machine can now produce. "AI did not create this obsolescence. It revealed it," he writes. In one small experiment, people who wrote essays with a chatbot showed weaker connections between brain regions while writing and remembered less of what they had produced. The study tracked 54 writers wearing brain-monitoring caps. The ones given AI showed the faintest mental engagement of any group and felt little ownership over their own essays. Schools that resist None of this explains why schools are so hard to change, which is the puzzle Zhao spends much of the paper on. Reformers have tried for a century, yet the basic structure holds. Classes are still sorted by age, subjects still stand apart, and tests and rankings still run the show. Researchers call this stubborn pattern "the grammar of schooling," the deep set of habits that outlasts reform after reform. Built to stay the same Zhao's own explanation for its staying power is a peace treaty. A school, he argues, is a settlement among people who want different things from it. Parents want clear signs their children are doing well. Colleges want familiar credentials for sorting applicants, and governments want numbers they can compare across schools. Grades, schedules, and standardized tests keep all these groups reasonably content at once, so any change that unsettles the arrangement meets resistance before it starts. Professor Zhao draws a line between improvement, which makes the existing system run better, and transformation, which questions what the system is for. The courageous minority Zhao's answer is not to fix the whole system at once, which he thinks is close to impossible. It is to start where change is actually within reach. In his words, a courageous minority is the small number of teachers, students, and leaders in almost any school who are unhappy with how things run. They rarely hold much power. What they do hold is a space they control, be it a single classroom, a small mentoring group, a capstone project, or a school-within-a-school. Professor Zhao builds on panarchy theory, a way of describing how complex systems change. The theory suggests that large systems rarely change from the top down. Small, protected corners are freer to experiment, and a workable experiment can spread outward, winning allies and slowly remaking the whole. Work that machines can't do In practice, this looks like ordinary teachers rebuilding ordinary assignments. Instead of a standard persuasive essay, students might pick a real local problem, gather their own data, and pitch recommendations to an audience that is not their teacher. A math class might study traffic around the school and propose safer crossings instead of solving invented rate problems. The point is to ask students to do work a machine cannot simply hand back. The theory is already being tested. Zhao and colleagues have started a network of about 20 schools across several countries, where educators join calls at odd hours to design small experiments inside their own buildings. These include student-directed learning days, student-run podcasts, and inquiry projects where children choose what is worth investigating. Changes for education Global labor forecasts expect AI to create around 170 million jobs and wipe out about 92 million by 2030, with the surviving work leaning on judgment and the ability to work well with others. A school that keeps rewarding routine output prepares students for the part of the economy shrinking fastest. Piling on more technology is not a safe default either. Population research has linked heavier screen time among children and teens to lower well-being, including weaker self-control and more trouble finishing tasks. The pattern held across thousands of young people in one large study. More digital tools do not automatically mean better learning. The future of learning with AI What Zhao adds to the debate is a change of target. The question is no longer how to keep AI out of school, but what school asks students to do once AI is in the room. His theory says real change will not arrive as a grand policy handed down from above. It will start with a few people building better learning in the spaces they already hold, and it may spread from there. The study is published in the journal ECNU Review of Education. -- - Like what you read? Subscribe to our newsletter for engaging articles, exclusive content, and the latest updates. Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.

Elite AI-powered schools insist they provide route to excellence

Long-term effects of incorporating the technology into learning are still unknown, but inequalities are emerging The newest entrant to New York's elite school market makes the ultimate investment pitch to parents: if your child does not graduate as a millionaire, you get your money back. Founders School in Manhattan promises to refund families its $150,000 annual tuition fee if their child has not made $1mn "in profit" upon completion of the four-year high school course. Inspired by its parent organisation, Alpha School, which is focused on children aged 5 to 18, Founders School aims to compress the school day into three hours of study, during which pupils are taught by an "AI tutor" that adapts to their abilities. This frees up the rest of the time to work with mentors to build a business that the students own, such as services agencies and consumer apps. But its offer of a launch pad into entrepreneurial stardom lies beyond the reach of most families. There are only 20 spots available in Founders School's first intake this September, and it sets ruthless expectations for its pupils who will generally enrol at age 14. "Students who don't put in the effort or who drag down the cohort culture will be removed," the Founders School website says. Alpha School has become a prominent adopter of AI in education, earning vocal support from figures such as hedge fund billionaire Bill Ackman on social media. Rather than conventional grades, Alpha's preferred metric is "mastery" of the subject -- pupils do not move on to the next lesson until they achieve 90 per cent understanding of the material. There has been scant evaluation of AI's long-term outcomes in education, Stanford researchers note. But as AI usage sweeps through schools globally, some families are worried about the risks of embracing it. Founded in New York, the campaign group Parents for AI Caution in Educational Spaces has urged the city to impose a moratorium on the use of generative AI in New York schools. Hundreds of New York artists have also demanded the same, calling AI an "inequity machine that subjects communities of colour to surveillance" that "amplifies racial and gender biases". Researchers also caution that AI use could deepen an educational divide, because better-resourced schools and leadership teams are more able to reap the gains of the technology. There is a risk that, when not done properly, AI implementation "could reinforce existing inequalities between schools and pupils", write Teach First chief executive James Toop and Matt Prebble, Accenture UK and Ireland chief executive, in a joint report. Long-term cognitive effects also preoccupy researchers. A recent large-scale study in Chinese secondary education found AI adoption raised homework scores by 18 per cent, but monthly exam scores and "high-stakes entrance-exam scores" fell by about a fifth. However, some proponents argue the technology could help level the playing field, giving children access to personalised tutoring that adapts to their pace of learning. John Dalton, who pioneered the UK's first "teacherless classroom" which also compresses academic study into three hours, says that his school David Game College particularly benefits from the 100 languages on its AI adaptive platform because about half of its pupils are from overseas. The annual tuition fee is £35,000 for non-UK pupils. Dalton says its bespoke platform, designed for a GCSE-level programme, has been built to their education-only needs. "If a kid asks who won against Colombia last night in the World Cup, it will say, 'I can't tell you that, let's get back to your mathematics,'" he adds. It's so frustrating for kids because it's either, like, you cheat and you get ahead or you don't cheat and fall behind He says that swapping out a human teacher for an AI one is a fundamental advantage because it builds knowledge in a "non-judgmental way". Children are instructed according to their ability and can ask a question as many times as they want, saving them the embarrassment of admitting they do not know something to the whole classroom, he says. Alpha is also seeking to appeal globally and co-founder Joe Liemandt plans to scale the software to reach a billion pupils. "The mastery-based programme works for anyone, even if you are a 'C' or 'D' student in math, as long as you're motivated," says Anna Davlantes, Alpha's spokesperson. Teachers are called "guides" and there is a ratio of one to every five students. Maddie Price, who completed secondary school this year at an Alpha institution in Austin, Texas, says that the AI-enabled model of learning means she can interact with the "real world" or industry and pursue the kinds of self-directed projects that many in her generation desperately crave. "You know, with [generative] AI now, it's so frustrating for kids because it's either, like, you cheat and you get ahead or you don't cheat and fall behind," she says. She says Alpha guided her to use AI critically and paired her with mentors and professional theatre producers to create a TikTok musical. Mike Lambert, global education director of Inspired Education Group, which is launching seven AI-enabled primary schools including one in central London next year, says the data gleaned from its personalised teaching platform identifies "where children need support from teachers or need to be stretched". Students, including those with special educational needs, are not "having to pretend they know something". The afternoon workshops, targeting collaboration and public speaking, which he describes as "hard skills", are designed to prepare children for the workplace. However, that vision of AI as a tool for inclusivity sits alongside a reality that its most enthusiastic adopters include some of the world's most expensive schools. Inspired Education, an international group of 125 private schools, promises parents exceptional results from its e-learning model, which will charge higher fees than the traditional schools in its network. Founder Nadim Nsouli openly admits that it is "not a model for everyone". "It's a premium product," he says.

Study finds AI boosted homework scores 18% -- then tanked exam results 20% | Fortune

Students using AI shortcuts for their homework may be sacrificing their test scores -- even years down the line. New research published by the Centre for Economic Policy Research found that among 26,811 Chinese students in grades seven through 12, AI adoption increased homework scores by 18% and cut down completion time by 30%. However, within six months, monthly exam scores decreased by 20%, and college entrance examples fell by 18% to 24% -- with scores reaching their worst after two years. Researchers from Stockholm University and University of Hong Kong pinpointed the type of student most likely to experience these diverging scores: those who "outsourced" homework, deploying AI to complete the homework accurately, but in little time. The poor test scores, both in the long- and short-term were driven by about 80% of these students. The emergence of this student profile not only exposes a dissonance in AI productivity versus actual productivity gains -- it fuels an argument some educators and experts have against the unfettered use of technology in education more broadly. "For students, completing these tasks efficiently is not the goal; learning from them is," researchers wrote. "Hence, the rapid diffusion of generative AI tools among students in recent years has created widespread concerns about their learning...Our findings show that generative AI, which is likely to become a prevalent technology for education, has a substantial negative impact on student learning." Gen Z's interest or ability to engage critically with educational materials is being increasingly questioned as the generation becomes synonymous with AI adoption -- and cheating at school. An Atlantic cover story has hinted at nothing less than the onset of a new Dark Ages, fueled by a "post-literate" younger demographic concerned more with quickly and conveniently intaking huge amounts of information, and less with savoring and digesting it, atrophying the ability to think critically. This research suggests that whatever the reason, the incentives to use AI to skip the act of learning are simply overwhelmingly powerful -- and a problem society is failing to grapple with. Why students turn to AI AI use in schools has proliferated globally, with 84% of U.S. high schools students reporting using the technology for homework, according to a CollegeBoard survey of more than 1,000 high schoolers. Along with greater adoption has come misuse of the technology. Jacob Shelley, an associate professor of health law at Western University, told Fortune in May he was convinced his students cheated on a final exam, including using AI, for one of his classes, with 8% getting a perfect school on the multiple choice section, only to struggle on the essay portion, submitting answers with content not in the curriculum. "The results were anomalous," Shelley said. "That just never happened in 20 years of teaching." But rather than blame students for turning to the technology in high stakes moments, Shelley said he understands why students would feel compelled to cheat. Tech leaders like Anthropic's Dario Amodei and OpenAI's Sam Altman are now walking back predictions of an AI job apocalypse, but anxiety around the future of work in the world of AI still lingers. Computer scientist Cal Newport called these premonitions "doom trolling," accusing tech companies of manufacturing a fatalistic narrative around AI. They appear to have had an impact on the generation preparing to enter the workforce: Almost 90% of graduates from the class of 2026 are worried AI or automation could replace entry-level jobs, according to job search platform Monster. While economic data has yet to show an impact from AI on the labor market or productivity, Shelley said his students still feel the pressure to use the technology or risk being left behind. "AI is going to replace them, at least a lot of them, and they know that, and we're pretending that it won't," he said. "I think they see through it. So students are responsible, but I don't really blame them here." The folly of the teaching machine It may be no surprise to experts like neuroscientist Jared Cooney Horvath why homework gains thanks to AI aren't translating to learning or exam performance. Horvath -- who wrote in a testimony to the U.S. Senate Committee on Commerce, Science, and Transportation about how test scores indicate Gen Z is the first generation to be less cognitively capable than their parents -- has long opposed educational technology, or EdTech. He argues there's more than 100 years of evidence indicating automation can hinder learning, beginning in 1924 with the invention of the "teaching machine" Ohio State University psychology professor Sidney Pressey. Students would answer questions that a machine would displace when fed a piece of paper, but when asked outside the device to generalize their knowledge, they were unable to. Three decades later, legendary behaviorist B.F. Skinner produced his own version of the machine based on Pressey's prototype, where students would press keys indicating the correct answer, at which point another question would appear. But despite more advanced technology behind the mechanism, it yielded the same results, leaving both psychologists to abandon the project before it was implemented in schools. In a letter to Skinner, Pressey conceded that while students had not mastered the subject matter; they had just mastered the machine. "The reason they all quit was the transfer problem," Horvath said. "They found that kids would be very good so long as they were using the tool, but as soon as they went off the tool, they couldn't do it anymore." AI learned has the potential to once again recreate the problems of the teaching machine, Horvath argued. While teachers have found some benefits to AI in the classroom -- such as scaffolding text to individual students' lexile levels, particularly English-language-learners -- Horvath has deja vu. AI can individualize learning by generating answers to specific queries, but it does not produce the friction or enable the critical thinning necessary for learning subject matters, he argued. "The tools experts use to make their lives easier are not the tools children should use to learn how to become experts," Horvath said. "When you use offloading tools that experts use to make their lives easier as a novice, as a student, you don't learn the skill. You simply learn dependency."

America's math and reading scores tanked after schools ditched textbooks for screens -- and AI could worsen the brain rot | Fortune

At the turn of the century, educational technology initiatives put laptop keyboards at the fingertips of U.S. schoolchildren. Now, 25 years later, the next generation of students have turned to AI -- and education experts warn unrestricted use of the technology could atrophy critical thinking skills. AI use among students has become ubiquitous following the 2022 release of ChatGPT. More than 80% of high school students reported using generative AI for schoolwork last year, according to a survey from College Board, the nonprofit that administers the SAT and AP tests. While access to AI chatbots makes homework as easy as plugging a question into one's phone, the frictionless retrieval of information using AI has raised concerns among educators: Rather than aid in learning, could AI actually hinder the process? A Brookings Institute study published in January laid bare anxieties about the potential harms of AI in the classroom. Analyzing data from interviews and focus groups with more than 500 educators, parents, and students across 50 countries, as well as from more than 400 studies, the researchers found at this point, "risks of utilizing generative AI in children's education overshadow its benefits." The report gave credence to early research -- including a February 2025 Microsoft study -- finding AI use was associated with worse judgement and critical thinking skills. "The cognitive offloading, and the cognitive decline that's associated with that, the decline in critical thinking, and just even reading and writing and knowledge of basic facts -- I absolutely believe that," to be the case, Mary Burns, an education consultant and co-author of the Brookings Institute study, told Fortune. Why is EdTech under scrutiny? Computer use in schools has come under recent scrutiny following a Congressional testimony in January from neuroscientist Jared Cooney Horvath, who noted, citing Program for International Student Assessment data, that Gen Z is the first generation in modern history to be less cognitively capable than their parents. He blamed unfettered access to classroom technology, noting a stark correlation in lower standardized testing scores and more screen time in school. A 2014 study surveying 3,000 university students found that two-thirds of the time students spend on their screens were on off-task activities. "This is not a debate about rejecting technology," Horvath said in his written testimony. "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." Horvath, author of the 2025 book The Digital Delusion: How Classroom Technology Harms Our Kids' Learning -- and How to Help Them Thrive Again, told Fortune the rise of EdTech was a result of tech companies creating a narrative around the need for screens in the classroom to bolster learning. The push for computers in schools began in 2002, when Maine became the first state to introduce a statewide program providing laptops to schoolchildren in the classroom. Following a slow rollout, Google began reaching out to educators to test its low-cost Chromebook with free Google apps, and asked teachers and administrators to promote the product. In partnership with schools, Google's Chromebook became commonplace in classrooms, accounting for more than half of digital devices sent to schools in 2017. There have been more than 100 years of evidence showing the failures of automated learning, Horvath argued, beginning with the 1924 invention of the "teaching machine" by Ohio State University psychology professor Sidney Pressey. Students learned to answer the questions the machine would generate when fed a piece of paper, but were unable to generalize that knowledge outside the device. "Kids would be very good so long as they were using the tool, but as soon as they went off the tool, they couldn't do it anymore," Horvath said. Burns, the education consultant, said AI was, in some ways, a natural extension of the argument tech companies have made about the need for computers in school, which is that students are able to learn at their own pace, or seek out information of interest to them to initiate their own learning. "[Tech] companies keep talking about, AI is personalizing learning," she said. "I don't think it's personalizing learning. I think it's individualizing learning. There's a difference there, and that's kind of a classic carryover from educational technology." How is AI being integrated into classrooms? According to Horvath, student AI use is not conducive to learning because it mirrors the failures of the 20th century "teaching machines." Students' learning was individualized -- they answered questions from the device at their own pace and independently from other students -- but were unable to synthesize knowledge taught outside the device. Similarly, Horvath said, giving AI to students without clear instructions or parameters teaches students how to rely on the device, not their own critical thinking. A University of Wisconsin survey published last month found that of 303 educators and school professionals in the state and 132 education professionals across the country, less than one-third reported implementing policies or guardrails around AI use. "The tools experts use to make their lives easier are not the tools children should use to learn how to become experts," Horvath said. "When you use offloading tools that experts use to make their lives easier as a novice, as a student, you don't learn the skill. You simply learn dependency." Burns -- a proponent of EdTech -- said it's futile to eschew the technology altogether. The Brookings Institute study found that despite educators having real fear that students will use AI to cheat, teachers are using AI to create lesson plans. Data on AI in the classroom is limited, but there are benefits, she added. For English language learners, for example, teachers can use AI to alter the lexile level of a reading passage. "To say that technologies are a failure is not true," Burns said. "To say technology is a mixed bag is true." A version of this story was published on Fortune.com on March 14, 2026. More on AI in schools:

AI labs begin to muscle in on $6tn education market

From the moment OpenAI's ChatGPT burst on to the scene in 2022, schools and universities have been playing catch-up. Many educators banned AI tools outright at first, but quickly saw that large language models had become a fact of life. Now, they are grappling with how to prepare their students for a world being transformed by the technology -- and the world's top AI builders are getting in on the action. Anthropic has just launched Claude for Teachers, a free tool aimed at K-12 (kindergarten to age 18) teachers in the US that aims to help them plan lessons. OpenAI has ChatGPT for Teachers -- a free, self-serve offer for verified US K-12 teachers and staff. It also offers ChatGPT Edu, a discounted enterprise subscription aimed at universities. Google, meanwhile, has a suite of tools based around its Gemini models targeting teachers and students, building on its existing cloud tools for educational institutions. For the world's biggest AI labs, education represents both a chance to help society adapt to the seismic changes their inventions are bringing, and an enormous business opportunity. A report by Morgan Stanley put the size of the global education market at $6tn in 2022. And in students, AI companies have the chance to sign up tomorrow's leaders and workers. "Let's not be naive, AI is absolutely a marketplace for these companies," says Simon Buckingham Shum, professor of learning informatics at University of Technology Sydney. Nonetheless, AI labs say there is an imperative for them to help ensure their tools aid rather than hinder future generations. "Every student today is going to grow up with AI . . . it's our job to help shape how students use the technology," says Leah Belsky, vice-president of education at OpenAI and a former chief revenue officer at Coursera, the online learning platform. The company is working with governments to bring its technology into the classroom. A partnership launched last year with Estonia has already brought tools such as ChatGPT Edu into the hands of over 30,000 university and secondary school students, teachers and researchers. The venture is part of OpenAI's Education for Countries programme that aims to "personalise learning" and "prepare students for the workforce". The AI lab is working with partners in Estonia, Greece, Jordan, Kazakhstan, Slovakia, Trinidad and Tobago, the United Arab Emirates and Italy. In each of these countries, says Belsky, the company plans to roll out customised versions of its chatbot, tailored to the local pedagogical style and built to "provoke curiosity". She and other education executives at AI labs acknowledge fears among teachers that the technology, with its instantaneous responses to prompts, could reduce the productive struggle that helps students learn. But they contend that AI, if used correctly, could enhance the learning experience. Both OpenAI and Anthropic espouse a "Socratic" approach to learning, turning chatbots into tutors that guide critical thinking. For example, Anthropic's "Claude for Education" -- intended for university teachers and students -- includes a learning mode that asks "how would you approach this problem?" rather than providing answers immediately. Drew Bent, Anthropic's education lead, says the company's approach is centred on "working with the educators who really understand the students best". Claude for Teachers gives K-12 teachers across the US free access to its Claude Cowork AI agent and is intended to help them plan lessons and assignments and provide automated reports on students. Edtechs are 'nervous [and] afraid that the labs are going to [make them] obsolete' The companies perhaps best placed to monetise AI in education are the likes of Microsoft and Google, which have longstanding relationships with many universities through their cloud services. Both companies have incorporated their AI tools -- Copilot and Gemini, respectively -- into their education offerings, with more powerful tools for higher-tier, paid plans. "You're already deeply embedded and incentivised to use those stacks," says one education executive, who asked not to be named, speaking about competitors. All of this could present a threat to education's incumbent digital businesses, the edtechs -- but so far, the sector has chosen to partner with the frontier labs building the big foundation AI models, rather than fight them. Coursera has teamed up with OpenAI to integrate the learning platform's content into ChatGPT. Instructure, which sells a so-called learning management system that hosts assessments and learning materials, is a partner with both OpenAI, and Anthropic's Claude for Education initiative. Edtechs are "nervous [and] afraid that the labs are going to [make them] obsolete", says Michael Feldstein, who writes about the sector and is chief strategy officer at the education consortium 1EdTech. However, he adds that the companies' "information on patterns of usage" in education gives them data the AI labs do not have. For all the effort, however, adoption of AI at the institutional level has been slow. More than 40 per cent of students in higher education said AI had not been integrated into any of their courses, according to a recent study by the Digital Education Council, a global consortium of universities, colleges and schools. Of those whose courses had made use of AI, 42 per cent said it had been only "somewhat helpful", the DEC said, while 24 per cent reported "limited learning benefit". There's an open question as to whether schools are going to want to pay out for top chatbots when adequate models will run on their device for free In the longer term, there are questions over the commercial viability of enterprise AI in education, says Feldstein. "There's an open question as to whether schools are going to want to pay out for top chatbots when adequate models will run on their device for free," he says, referring to the growing popularity among businesses of free, so-called open-weight AI models, whose parameters are released publicly, meaning they can be fine-tuned easily for specific uses. At the University of Technology Sydney, Buckingham Shum says instructors are building their own AI products, with an internal quality assurance process to ensure the technology is suitable for students. "It's good that big tech companies are responding" to the disruption caused by AI, he says. "But in my view, the universities that are doing their jobs are putting design tools in the hands of educators."

Is AI killing critical thinking in the classroom?

At Thames Christian School in London, Madeleine Champagnie is less worried about her students overusing artificial intelligence than about them failing to use it enough. Despite widespread fears that young people are outsourcing their thinking and learning, she sees a reluctance to engage with technology's rising role as a risk to their long-term success. "Students are not all rushing to cheat with AI," says Champagnie, the school's innovation lead and head of English. "That's been a big misunderstanding. Loads of them don't want to use it. They are not always trying to get around the system. They are as scared of cognitive offloading as us," she says, referring to the idea that delegating thinking and research to technology hinders learning. "I've had to persuade some extremely bright students to try it," she adds. However, there is increasing evidence that not all educators are facing problems with that reticence. A number of research studies demonstrate what many teachers now fear: that students are increasingly relying on AI in their learning and, in the process, using it as a substitute for developing their own critical thinking skills and independent judgment. There are some controls on such excessive cognitive offloading in schools, notably the continued discipline of written "high-stakes" timed and invigilated exams without access to digital devices to test independent understanding. But some research has already shown that while AI use can boost short-term speed and results, its subsequent removal can leave students performing less well than those who never had it in the first place. Several educators are now experimenting with the idea that, when used with suitable safeguards, integrating the technology is not only necessary as the workplace adopts AI more broadly, but also offers the potential to boost learning. Steve Marshall-Taylor, headmaster of Brighton College, which provides pupils with access to a variety of AI tools, says he is "very positive" about the technology on condition "there are some guardrails". He stresses: "We have to make sure students can work with pen and paper for the English exam system." A report looking into cognitive offloading published in March by the Australian Network for Quality Digital Education cautions against the dangers of the "false mastery" by students using AI that results in detrimental offloading or "cognitive atrophy". But it also suggests there is value in the "beneficial offloading" of lower-order tasks to AI to free learners to focus on essential intrinsic tasks. The approach can help foster "deep learning", the authors say. For example, the technology can be used to check grammar and syntax, liberating students in humanities to evaluate evidence, structure arguments and synthesise sources. In maths, it can provide worked examples and tests of understanding of a quadratic formula, rather than simply generating the answer to a set question. We need to unpick the old processes and identify which bits matter "Teachers are still in charge of overall learning," says Kate Erricker, interim chief education officer at Nord Anglia, an international network of private schools. "They set the tone. AI can then personalise to each individual student's area of need." She has found the technology has been particularly valuable in supporting non-native English language speakers. The personalised tutoring can include the use of so-called Socratic dialogue in which the AI tool poses questions that guide a student through a piece of learning, rather than provide answers. It offers scope for students to progress at their own pace, without embarrassing them, distracting others or consuming teachers' time. Elsewhere, AI is being adapted to go beyond written text questions and answers. Dan Wang, a professor at Columbia University Business School, has developed CAiSEY, which holds conversations with students as they prepare business teaching case studies ahead of classroom discussion. He has found the tool helps them explore a wider range of ideas, come better prepared and be more engaged in class. It also aids the professor in analysing the students' initial ideas to help better structure the final group discussion. Efekta Education, an edtech company, has developed an AI platform that uses digital avatars backed by AI to help personalise learning. It claims to have helped students to achieve a sharp rise in English proficiency in countries including Brazil and Rwanda. More broadly, Rose Luckin, professor emerita at University College London and chief executive of the consultancy Educate Ventures Research, argues that it may be time to rethink the very process of education -- including assessment. "We need to verify what students have understood, and how you build critical thinking and a sophisticated learning capability," she says. "We need to unpick the old processes and identify which bits matter."