AI in Education

AI in education refers to the use of artificial intelligence technologies to support teaching, learning, assessment, and administrative processes across educational settings. Applications range from intelligent tutoring systems and automated grading to personalized learning platforms and AI-powered administrative tools. The field traces its roots to the 1960s and 1970s, with early computer-assisted instruction systems like PLATO, and has expanded rapidly since the release of large language models such as ChatGPT in late 2022. As of 2026, AI tools are embedded in classrooms from primary schools through universities, reshaping how students learn and how educators teach, while also raising serious questions about academic integrity, equity, and data privacy.

The core promise driving the field is scaled personalization. In a landmark 1984 paper, educational psychologist Benjamin Bloom found that students given one-to-one tutoring plus mastery learning scored roughly two standard deviations (two sigma) higher than students in a conventional classroom, with the average tutored student outperforming about 98% of conventionally taught peers ^[15]. Because human one-to-one tutoring is too expensive to provide to every student, Bloom framed this as the "2 sigma problem," and modern AI tutors are widely positioned as the first scalable attempt to close that gap. By early 2026 the global AI in education market was estimated at $7 billion to $19 billion and an estimated 86% of higher education students reported using AI as a primary research and brainstorming partner ^[9][6].

What is the history of AI in education?

Early computer-assisted instruction (1960s-1970s)

The earliest attempts to use computers for education date to the late 1950s and 1960s. The most significant of these was PLATO (Programmed Logic for Automated Teaching Operations), created in 1960 by Donald L. Bitzer at the University of Illinois at Urbana-Champaign. PLATO was the first generalized computer-assisted instruction system, allowing educators to create interactive lessons that students could work through at their own pace on networked terminals ^[1].

PLATO was remarkably ahead of its time. The system introduced several innovations that would not become mainstream for decades, including a plasma display, touch screen input, graphical user interfaces, message boards, chat rooms, instant messaging, and even real-time multiplayer games. The courseware was written in TUTOR, a programming language conceived in 1967 by biology graduate student Paul Tenczar that allowed non-programmers to design lesson modules ^[1].

By the mid-1970s, PLATO IV supported over 900 terminals at 146 sites across the United States and Canada. Control Data Corporation (CDC) licensed the system from UIUC and began commercializing it, and by the mid-1980s over 100 PLATO systems were operating worldwide, mostly at educational institutions ^[1]. Although PLATO eventually faded from use, its influence on educational technology was substantial: it demonstrated that computers could deliver interactive, individualized instruction at scale.

Bloom's 2 sigma problem (1984)

The scientific case for personalized instruction was crystallized in 1984 when Benjamin S. Bloom published "The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring" in Educational Researcher (volume 13, issue 6, pages 4-16). Drawing on doctoral studies by his University of Chicago students Joanne Anania and Joseph Arthur Burke, Bloom compared conventional classroom instruction, group mastery learning, and one-to-one tutoring. He reported that the average tutored student performed about two standard deviations better than the average student in a conventional class of roughly 30, placing the tutored student above 98% of the comparison group ^[15].

Bloom called this a "problem" rather than a solution because human one-to-one tutoring is far too costly to provide to every student, and he challenged researchers to find scalable group methods that could match it. Four decades later, this framing has become the dominant rationale for AI in education. AI tutoring products are routinely pitched as the technology that can finally deliver tutor-like personalization at near-zero marginal cost, turning what Bloom called the 2 sigma problem into a 2 sigma opportunity. The strength of the original effect is debated: some researchers note the small sample sizes and short interventions in the underlying studies, and large-scale replications have generally found smaller, though still positive, tutoring effects ^[15].

Intelligent tutoring systems (1970s-1990s)

While PLATO and similar systems delivered pre-programmed instruction, researchers in artificial intelligence pursued a different approach: building systems that could adapt their teaching strategy based on a model of the student's knowledge. These came to be known as Intelligent Tutoring Systems (ITS).

Jaime Carbonell's SCHOLAR system (1970) was among the first to use AI techniques for instruction, employing a semantic network to engage students in Socratic dialogue about South American geography. Other early ITS projects included WHY (Collins and Stevens, 1977), which explored reasoning strategies, and SOPHIE (Brown, Burton, and de Kleer, 1982), which taught electronic troubleshooting ^[2].

The most commercially successful ITS of this era was Carnegie Learning's Cognitive Tutor, developed in the 1990s based on John Anderson's ACT-R cognitive architecture at Carnegie Mellon University. The Cognitive Tutor for algebra tracked individual student mastery of specific skills and adjusted problem selection accordingly. A large-scale randomized controlled trial published in the Journal of Research on Mathematics Education found that students using Cognitive Tutor outperformed control groups on standardized tests ^[2].

Despite promising results, ITS adoption remained limited through the 1990s and 2000s due to high development costs, the need for extensive domain modeling, and the difficulty of scaling these systems beyond narrow subject areas.

Rise of adaptive learning platforms (2010s)

The 2010s saw the emergence of commercial adaptive learning platforms powered by machine learning rather than hand-crafted expert rules. Companies like Knewton (founded 2008), DreamBox Learning (founded 2006), and Carnegie Learning (which continued to evolve its Cognitive Tutor platform) used data-driven algorithms to personalize content sequencing, pacing, and difficulty.

Knewton, which raised over $180 million in venture funding, partnered with major publishers including Pearson to embed its adaptive engine into digital textbooks. The company claimed its technology could identify precisely which concepts a student had mastered and which needed reinforcement, adjusting the learning path in real time. However, Knewton struggled commercially and was acquired by Wiley in 2019 for a fraction of its peak valuation, illustrating the gap between the promise and the business reality of adaptive learning.

The Massive Open Online Course (MOOC) movement, led by platforms like Coursera and edX (both launched in 2012), also brought AI into education at scale, using machine learning for course recommendations, automated grading of programming assignments, and peer review matching.

The LLM turning point (2022-present)

The release of ChatGPT in November 2022 changed the landscape dramatically. For the first time, students and teachers had access to a general-purpose AI that could write essays, solve math problems, explain complex concepts, generate lesson plans, and engage in open-ended dialogue. Within two months, ChatGPT reached 100 million users, and education was one of the sectors most immediately affected ^[3].

The initial reaction from educational institutions was largely one of alarm. Several school districts, including New York City Public Schools, banned ChatGPT on school networks in January 2023, citing concerns about academic integrity. However, most of these bans were reversed within months as educators recognized that prohibition was neither practical nor desirable. By mid-2023, the conversation had shifted from "how do we block this" to "how do we integrate this thoughtfully" ^[3].

The optimistic case for AI in education was articulated forcefully by Khan Academy founder Sal Khan in his April 2023 TED talk, "How AI could save (not destroy) education." Khan argued, "We're at the cusp of using AI for probably the biggest positive transformation that education has ever seen," framing AI tutors as a way to finally give every student a personal tutor at scale and address Bloom's 2 sigma problem ^[16].

How is AI used in classrooms today?

AI is now applied across a wide range of educational contexts. The following table summarizes the major application areas as of 2026:

Scroll sideways for more →

Khan Academy Khanmigo

Khan Academy was among the first educational organizations to partner with OpenAI to build an AI tutoring product. Khanmigo, launched in 2023 and powered by GPT-4, functions as a personal tutor and teaching assistant. Unlike general-purpose chatbots, Khanmigo is designed to guide students toward answers rather than simply providing them. When a student asks for help with a math problem, the system asks probing questions and offers hints, mimicking the Socratic method ^[4].

Khanmigo expanded rapidly during the 2024-25 school year, growing from 40,000 to 700,000 K-12 students, with projections to surpass one million in the 2025-26 academic year. The tool is now used across hundreds of school districts in the United States and has begun reaching classrooms in India, Brazil, and the Philippines. For individual users, Khanmigo costs $4 per month or $44 annually, while teacher-support tools are offered for free ^[4].

Khanmigo covers math, science, reading, writing, computer science, and test preparation. For teachers, it can generate lesson plans, create assessments aligned to specific standards, and provide insights into student progress.

Duolingo Max

Duolingo, the language learning platform with over 100 million monthly active users, launched Duolingo Max in March 2023 as a premium tier powered by GPT-4. The subscription introduced two AI-driven features: Roleplay, which places learners in simulated real-world conversations (ordering coffee in a Parisian cafe, negotiating at a market), and Explain My Answer, which provides detailed, contextual feedback on why an answer was correct or incorrect ^[5].

In 2024, Duolingo added Video Call with Lily, enabling users to have voice conversations with an AI character, providing speaking practice that was previously difficult to access without a human conversation partner. Internal data from Duolingo shows 34% better grammar retention with the Explain My Answer feature, and 78% of Roleplay users report increased speaking confidence within four weeks ^[5].

By 2025, Duolingo had used AI to launch 148 new language courses, dramatically accelerating content production through AI-assisted creation with human oversight. Duolingo Max is priced at $29.99 per month or $167.99 annually ^[5].

MagicSchool AI and teacher-focused platforms

MagicSchool AI has emerged as one of the most widely adopted AI platforms designed specifically for educators. By 2025, it reported over 5 million educator users. The platform provides over 60 AI-powered tools tailored to teacher workflows, including:

Scroll sideways for more →

Other teacher-focused AI platforms include Eduaide.AI, SchoolAI, and Brisk Teaching. These tools aim to save teachers time on routine tasks, with surveys finding that teachers who use AI tools at least weekly save an average of 5.9 hours per week, adding up to roughly six extra weeks of reclaimed time across a standard school year ^[10].

How do students and teachers use LLMs in education?

Student use of ChatGPT and other LLMs

Student adoption of generative AI tools has been swift and continues to accelerate. A Higher Education Policy Institute (HEPI) survey conducted in February 2025 found that 88% of students reported using generative AI tools such as ChatGPT for assessments. Pew Research Center data from January 2025 showed that 26% of U.S. teens already use ChatGPT for schoolwork, and a September 2025 RAND survey found that 54% of K-12 students used AI for school, up more than 15 percentage points in two years. By early 2026, an estimated 86% of all students in higher education utilize AI as their primary research and brainstorming partner ^[6][10].

Students use LLMs for a variety of academic tasks: brainstorming essay topics, getting explanations of complex concepts, debugging code, summarizing readings, practicing for exams, and generating first drafts. Some of these uses are clearly constructive (using an LLM as a study aid is analogous to asking a knowledgeable friend for help), while others cross into academic dishonesty (submitting AI-generated text as one's own work).

The line between legitimate use and misuse is not always clear, and it varies by institution, course, and assignment. This ambiguity has been one of the central challenges for educators since 2023.

Teacher and instructor tools

LLMs have also become popular tools for educators themselves. Teachers use AI to generate lesson plans, create differentiated materials for students at different levels, write quiz questions, provide feedback on student drafts, and handle administrative tasks like writing recommendation letters or parent communications.

Platforms like MagicSchool AI, which reported over 5 million educator users by 2025, provide teacher-specific AI tools with built-in guardrails. These platforms are designed to generate content aligned to educational standards (such as Common Core or Next Generation Science Standards) and to save teachers time on routine tasks, potentially freeing them to spend more time on direct instruction and student interaction.

How is AI being implemented in K-12 schools?

The integration of AI into K-12 education has progressed rapidly, though unevenly, across the United States and globally.

State-level AI guidance

As of March 2025, 28 states have published or adopted AI guidance for K-12 education, up from fewer than 10 in early 2024. These guidelines vary significantly in scope and specificity:

Scroll sideways for more →

Approximately 74% of districts planned to provide teacher AI training by fall 2025, reflecting growing recognition that effective AI integration requires teacher preparation ^[10].

Teacher readiness gap

A significant challenge exists in educator preparedness. While 63% of U.S. teens report using AI tools like ChatGPT for schoolwork, only 30% of teachers report feeling confident using these same AI tools. Student use of AI for school-related purposes jumped 26% since the previous school year, while educator use rose 21% over the same period ^[10].

This readiness gap creates situations where students may be more familiar with AI capabilities and limitations than their teachers, complicating classroom management and assessment design.

Does AI actually improve learning outcomes?

Research on AI-powered personalized learning has produced increasingly strong evidence of effectiveness, though with important caveats. Much of this research is framed explicitly against Bloom's 2 sigma benchmark, asking whether AI tutoring can approach the gains historically seen only with human one-to-one tutoring ^[15].

Harvard AI tutoring study

A 2024 study by researchers at Harvard found that students using an AI tutor in a computer science course learned 1.5 times faster and scored significantly higher on assessments than those in a traditional lecture format. The study controlled for prior knowledge and motivation, suggesting that the AI tutor's ability to provide immediate, individualized feedback was a genuine advantage ^[8].

A 2025 Harvard University physics study produced even more striking results, finding that students using AI tutors learned more than twice as much in less time compared to those in traditional active-learning classrooms ^[10].

Meta-analysis of AI-assisted learning

A meta-analysis published in Computers & Education in 2025 reviewed 83 studies of AI-assisted learning and found a moderate positive effect (Cohen's d = 0.41) on learning outcomes compared to traditional instruction. The effect was strongest for personalized tutoring systems and weakest for simple content recommendation engines ^[8].

Khan Academy engagement data

Khan Academy's internal research has shown that students using Khanmigo spend 30% more time on task compared to students using the platform without the AI tutor, suggesting that the conversational interface increases engagement ^[4].

Duolingo learning outcomes

In language learning, Duolingo's data indicates that the Roleplay feature in Duolingo Max leads to measurable improvements in conversational confidence and grammar retention. The company reports that users who engage with Roleplay sessions at least three times per week show 34% better grammar retention on follow-up assessments ^[5].

Limitations of current research

Despite these promising findings, 86% of educational organizations use generative AI, the highest of any industry in the United States, yet most AI-based educational tools have not undergone independent validation, and few have been tested through rigorous methods such as randomized controlled trials. The gap between adoption rates and scientific validation remains a significant concern ^[10].

How is AI changing assessment?

AI is transforming both how students are assessed and how assessments are created.

AI-powered assessment creation

Teachers increasingly use AI to generate assessments. AI tools can create multiple-choice questions, short-answer prompts, and essay questions aligned to specific learning standards. More advanced tools can generate assessments at multiple difficulty levels, create alternative versions of the same test to reduce cheating, and produce detailed rubrics.

Automated essay scoring

Automated essay scoring (AES) systems use NLP to evaluate student writing on dimensions including organization, vocabulary, grammar, and argument quality. The Educational Testing Service (ETS) has used its e-rater system as a component of GRE essay scoring for over two decades. Newer AI-powered systems from companies like Turnitin (through its Gradescope platform) can provide more detailed, formative feedback.

However, AES systems have known limitations:

• They tend to favor longer essays and more complex vocabulary, regardless of argument quality
• They can be gamed by students who write grammatically correct but semantically empty text
• They may disadvantage non-native English speakers and students from different cultural backgrounds
• They struggle with creative or unconventional writing styles

Assessment redesign

Schools that have redesigned assessments to be more AI-resilient, moving toward oral exams, in-class writing, process portfolios, and project-based evaluations, report 40% fewer AI-related integrity issues compared to those relying solely on detection ^[6].

What are the risks and challenges of AI in education?

Academic integrity and cheating

The most immediate and widely discussed concern about AI in education is its impact on academic integrity. AI-related misconduct grew from 1.6 to 7.5 cases per 1,000 students between 2022 and 2026. Over 8 in 10 students reported using generative AI during the school year in 2025, and for many, that included graded assignments ^[6].

The detection problem is severe. While 68% of teachers report using AI detection tools, only 54% of faculty feel effective at identifying AI-generated content. Research from the University of Reading found that 94% of AI-generated submissions went undetected in a blind evaluation. This creates an uneven playing field: students who use AI without disclosure may receive higher grades than those who do their own work ^[6].

Three in four chief technology officers at educational institutions say that artificial intelligence has proven to be a moderate or significant risk to academic integrity at their institution ^[6].

Bias in AI detection tools

AI detection tools themselves have introduced a new form of inequity. Studies have found that non-native English speakers face a 61.2% false positive rate (meaning their original work is flagged as AI-generated), compared to just 5.1% for native speakers. This disparity has led institutions including Princeton and MIT to advise against relying solely on AI detectors ^[6].

The bias arises because non-native speakers sometimes produce writing that is more formulaic or uses simpler sentence structures, patterns that detection algorithms associate with AI generation. As a result, the students most likely to be falsely accused of cheating are often those who are already marginalized in educational settings.

Over-reliance on AI

Educators worry that students who rely heavily on AI tools may fail to develop critical thinking, writing, and problem-solving skills. If a student can get an AI to write every essay, they may never learn to organize their thoughts, construct an argument, or revise their own prose. This concern is particularly acute in foundational courses where the process of struggling with material is itself the learning objective.

Similar concerns have been raised about math and science education. When students use AI to solve problems step by step, they may get correct answers without understanding the underlying concepts. The risk is that AI becomes a crutch rather than a scaffold.

Data privacy

AI educational tools collect substantial amounts of student data, including performance metrics, learning patterns, interaction logs, and sometimes biometric data (in the case of proctoring tools). This data collection raises privacy concerns, particularly for minor students. In the United States, the Family Educational Rights and Privacy Act (FERPA) and the Children's Online Privacy Protection Act (COPPA) govern student data, but the application of these laws to AI tools is not always clear.

The European Union's General Data Protection Regulation (GDPR) imposes stricter requirements, particularly regarding automated decision-making and the processing of children's data. Schools using AI tools must ensure compliance with these regulations, which can be complex when tools are provided by third-party vendors whose data practices may not be fully transparent.

The digital divide

AI tools in education risk widening existing inequalities. Students at well-funded schools with strong technology infrastructure and institutional AI subscriptions have access to tools that students at under-resourced schools do not. Paid tiers of AI tools (like Duolingo Max at $29.99/month or ChatGPT Plus at $20/month) are out of reach for many families.

This digital divide extends globally. While schools in North America and Western Europe are debating how to integrate AI into curricula, schools in many developing countries lack reliable internet access, let alone AI-powered learning platforms. Without deliberate efforts to ensure equitable access, AI in education could become another mechanism through which privilege compounds.

How are schools and universities responding?

University and school policies

Educational institutions have taken varied approaches to AI use. A 2025 analysis of 174 universities found that only 30% had an explicit AI policy, while the remaining 70% had no AI-specific guidelines. Among those with policies, approaches range from broad permission with disclosure requirements to strict restrictions on summative assessments. Two-thirds of higher education institutions have or are developing guidance on AI use ^[7][10].

Common policy elements include:

Scroll sideways for more →

The trend as of 2026 is moving toward course-level discretion with disclosure requirements. Institutions increasingly recognize that a one-size-fits-all policy is impractical given the diversity of disciplines and assignment types ^[7].

AI detection tools

Several tools have been developed to detect AI-generated text, including Turnitin's AI detection feature (integrated into its plagiarism detection platform), GPTZero, Originality.ai, and Copyleaks. However, their reliability remains contested.

Duke University does not recommend AI detection software as part of an AI policy because the products are unreliable. OpenAI itself withdrew its own detection tool (AI Text Classifier) in July 2023 due to low accuracy. Australia's higher education regulator TEQSA warned in 2025 that AI-assisted cheating is "all but impossible" to detect consistently, and urged universities to redesign assessments rather than depend on AI detectors ^[7].

Schools that have redesigned assessments, moving toward oral exams, in-class writing, process portfolios, and project-based evaluations, report 40% fewer AI-related integrity issues compared to those relying solely on detection ^[6].

AI literacy curricula

A growing number of schools and districts are developing AI literacy curricula as a core competency. These programs teach students:

• How AI tools work at a conceptual level (training data, pattern recognition, probabilistic outputs)
• How to evaluate AI-generated content critically (checking for accuracy, identifying potential biases)
• Effective prompting techniques for productive AI use
• Ethical considerations surrounding AI (privacy, bias, labor displacement, environmental impact)
• The limitations of AI (hallucinations, lack of true understanding, potential for reinforcing stereotypes)

Several organizations have developed AI literacy frameworks, including ISTE (International Society for Technology in Education), AI4K12 (a national initiative for K-12 AI education), and UNESCO's AI competency framework for students ^[10].

What does the research say about positive impacts?

Despite the concerns, research has documented several positive effects of AI in education:

A 2024 study by researchers at Harvard found that students using an AI tutor in a computer science course learned 1.5 times faster and scored significantly higher on assessments than those in a traditional lecture format. The study controlled for prior knowledge and motivation, suggesting that the AI tutor's ability to provide immediate, individualized feedback was a genuine advantage ^[8].

Khan Academy's internal research has shown that students using Khanmigo spend 30% more time on task compared to students using the platform without the AI tutor, suggesting that the conversational interface increases engagement ^[4].

In language learning, Duolingo's data indicates that the Roleplay feature in Duolingo Max leads to measurable improvements in conversational confidence and grammar retention. The company reports that users who engage with Roleplay sessions at least three times per week show 34% better grammar retention on follow-up assessments ^[5].

A meta-analysis published in Computers & Education in 2025 reviewed 83 studies of AI-assisted learning and found a moderate positive effect (Cohen's d = 0.41) on learning outcomes compared to traditional instruction. The effect was strongest for personalized tutoring systems and weakest for simple content recommendation engines ^[8].

How big is the AI in education market?

The AI in education market has grown rapidly and projections vary across research firms:

Scroll sideways for more →

The wide range of market estimates reflects differences in how firms define the boundaries of "AI in education" (some include only pure-play AI EdTech, while others include AI features within broader EdTech platforms). Regardless of the exact figures, the growth trajectory is clear ^[9].

North America dominates the market with approximately 38% share in 2025, driven by high EdTech investment and widespread institutional adoption. Asia Pacific is the fastest-growing region, with countries like India, China, and South Korea investing heavily in AI-powered educational infrastructure ^[9].

What is the current state of AI in education (2025-2026)?

As of early 2026, AI in education is in a period of rapid, sometimes chaotic integration. Several trends define this moment:

Normalization of AI use. The initial shock and resistance that followed ChatGPT's launch has given way to a more pragmatic approach. Most institutions now accept that students will use AI tools and are focusing on how to channel that use productively rather than trying to prevent it entirely.

Assessment redesign. The most forward-thinking institutions are redesigning assessments to be more AI-resilient. This includes greater emphasis on oral examinations, in-class writing under supervised conditions, process portfolios that document the evolution of student work, and project-based assessments that require demonstrating competence in real time. Schools taking this approach report significantly fewer integrity issues ^[6].

AI literacy as a learning objective. A growing number of schools are treating AI literacy as a core competency, teaching students not only how to use AI tools effectively but also how to evaluate their outputs critically, understand their limitations, and consider the ethical implications of their use. Twenty-eight states have published K-12 AI guidance, and organizations like AI4K12 are developing standardized frameworks for AI education.

Expansion of AI tutoring. AI tutoring products like Khanmigo are scaling beyond pilot programs to widespread adoption. Khan Academy's growth from 40,000 to 700,000 student users in a single academic year suggests that the demand for AI-powered tutoring is substantial ^[4].

Teacher augmentation over replacement. The discourse has shifted away from fears that AI will replace teachers. The emerging consensus is that AI is most effective as a tool that augments teacher capabilities, handling routine tasks like generating practice problems, providing first-pass feedback on drafts, and answering common student questions, while freeing teachers to focus on the relational and motivational aspects of teaching that AI cannot replicate. Teachers using AI tools at least weekly save an average of 5.9 hours per week.

K-12 policy development. State-level AI guidance for K-12 education is expanding rapidly, with 28 states having published guidance by March 2025. As of the 2026 legislative session, MultiState was tracking 134 bills across 31 states related to AI in education, reflecting a sharp acceleration in lawmaker attention to the issue ^[13]. Idaho's S.B. 1227, for example, directs the state education department to develop a comprehensive generative AI framework covering privacy, procurement safeguards, transparency, academic integrity, and professional development. Pilot programs like Connecticut's are providing models for structured AI integration in public schools.

Higher education mandates. The State University of New York (SUNY) system announced in early 2026 that it will embed AI literacy into general education requirements for all incoming undergraduates starting Fall 2026, one of the first systems-wide mandates of its kind at a major public university ^[13].

OECD research on learning outcomes. The OECD Digital Education Outlook 2026, published in January 2026, issued a cautionary finding that has shaped institutional debate: while students using general-purpose generative AI tools often produce higher-quality outputs, these advantages tend to disappear, or even reverse, when AI access is removed in exams. The report concluded that AI in education is fundamentally a pedagogical challenge rather than a technology one, arguing that benefits depend on how AI is embedded in instructional design rather than on access alone. It also found that 37% of lower secondary teachers reported using AI in their work in 2024, with more than half using it to write or improve lesson plans, but that 72% expressed concerns about academic integrity ^[14].

Persistent equity gaps. Despite the potential of AI to democratize education, access remains uneven. Well-resourced schools and families benefit disproportionately, and the digital divide continues to be a limiting factor for equitable AI adoption in education worldwide.

The field is evolving rapidly, and the policies, tools, and pedagogical approaches of 2026 will likely look quite different from those of 2024. What seems clear is that AI is not a passing trend in education; it is a permanent feature of the learning landscape that institutions, educators, and students are still learning how to navigate.

See also

• Natural language processing
• AI ethics
• Generative AI
• GPT-4
• Intelligent tutoring system

References

1. "PLATO (computer system)." Wikipedia. https://en.wikipedia.org/wiki/PLATO_(computer_system) ↩
2. "Intelligent tutoring system." Wikipedia. https://en.wikipedia.org/wiki/Intelligent_tutoring_system ↩
3. "AI Cheating in Schools: 2026 Global Trends & Bias Risks." AllAboutAI. https://www.allaboutai.com/resources/ai-statistics/ai-cheating-in-schools/ ↩
4. "Meet Khanmigo: Khan Academy's AI-powered teaching assistant & tutor." Khan Academy. https://www.khanmigo.ai/ ↩
5. "Duolingo Max Uses OpenAI's GPT-4 For New Learning Features." Duolingo Blog. https://blog.duolingo.com/duolingo-max/ ↩
6. "75 AI in Education Statistics 2026 (Global Trends & Facts)." DemandSage. https://www.demandsage.com/ai-in-education-statistics/ ↩
7. "Generative AI Policies at the World's Top Universities: October 2025 Update." Thesify. https://www.thesify.ai/blog/gen-ai-policies-update-2025 ↩
8. "25 AI in Education Statistics to Guide Your Learning Strategy in 2026." Engageli. https://www.engageli.com/blog/ai-in-education-statistics ↩
9. "AI in Education Market Size to Surge USD 136.79 Bn by 2035." Precedence Research. https://www.precedenceresearch.com/ai-in-education-market ↩
10. "AI Pilot Programs in K-12 Settings." Education Commission of the States. https://www.ecs.org/ai-artificial-intelligence-pilots-k12-schools/ ↩
11. "U.S. Department of Education Issues Guidance on Artificial Intelligence Use in Schools." U.S. Department of Education, 2025. https://www.ed.gov/about/news/press-release/us-department-of-education-issues-guidance-artificial-intelligence-use-schools-proposes-additional-supplemental-priority
12. "AI Has Come for K-12 Education." American Enterprise Institute. https://www.aei.org/education/ai-has-come-for-k-12-education/
13. "AI in Education Legislation: 2026 State Policy Trends." MultiState, April 2026. https://www.multistate.us/insider/2026/4/9/how-states-are-regulating-ai-in-education-this-legislative-session ↩
14. "OECD Digital Education Outlook 2026: How generative AI can support learning when used with purpose." OECD, January 2026. https://www.oecd.org/en/publications/oecd-digital-education-outlook-2026_062a7394-en.html ↩
15. Bloom, Benjamin S. "The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring." Educational Researcher, vol. 13, no. 6, 1984, pp. 4-16. https://journals.sagepub.com/doi/10.3102/0013189X013006004 ↩
16. Khan, Sal. "How AI could save (not destroy) education." TED, April 2023. https://www.ted.com/talks/sal_khan_how_ai_could_save_not_destroy_education ↩