The Classroom That Changed While the Debate Continued
The education sector’s response to AI has followed a familiar pattern: initial alarm (ChatGPT was banned in New York City public schools in late 2022 within weeks of launch), followed by partial reversal (NYC lifted the ban and created guidance for AI use in May 2023), followed by gradual integration as educators moved from reactive policy to deliberate strategy. By 2026, the question of whether students and teachers will use AI is settled — they are, widely, and with varying degrees of institutional support. The more useful questions are which uses are pedagogically sound, which create risks, and how institutions are responding to a tool that arrived faster than policy could follow.
The answer to these questions looks different at different educational levels (primary school AI use raises different concerns than graduate-level research use), in different subject areas (AI writing assistance in language arts raises different questions than AI assistance in mathematics), and for different users (student AI use for completing assignments is governed by different considerations than teacher AI use for lesson planning).
How Teachers Are Using AI
Teacher AI adoption has focused on the tasks that consume disproportionate time without being the core of teaching craft: lesson plan generation (describing a learning objective and getting a structured lesson plan draft), differentiated material creation (generating versions of the same content at different reading levels for students with different needs), rubric and assessment design, parent communication drafting, and administrative document completion. These uses leverage AI’s strength in producing structured first drafts while keeping the teacher’s pedagogical judgment and knowledge of specific students in control of what’s actually used.
The AI tutoring application is the most transformative potential in education technology: an AI tutor that provides personalised, patient, unlimited explanations at a student’s specific level and pace addresses one of education’s most persistent constraints, which is that teacher time is finite while student learning needs are variable and individual. Khan Academy’s Khanmigo (a tutoring AI built on GPT-4) represents the early implementation of this concept; the evidence for its effectiveness relative to human tutoring is early but directionally positive for the use cases studied.
The Academic Integrity Question
The concern that generated the most immediate institutional response — students using AI to complete assignments that should represent their own work — is real, persistent, and not fully resolved by either detection tools or policy changes. AI detection software (Turnitin’s AI detection, GPTZero, Copyleaks) has documented false positive rates that make it unsuitable as an evidence standard for academic consequences; students who write authentically are sometimes flagged, and students who carefully edit AI output may not be. Policy approaches (prohibiting AI use, requiring disclosure, permitting AI as a tool) vary by institution and have different levels of enforceability.
The pedagogical response that many educators have found more productive than detection-and-punishment: redesigning assessments in ways that make AI completion less useful or impossible. In-class writing rather than take-home essays; oral presentations that require explaining reasoning; assignments that require personal experience or local knowledge that AI doesn’t have; iterative assignments where each step builds on specific prior work; and process portfolios that document thinking over time rather than just final products — these approaches assess what AI can’t fake rather than trying to prevent students from using AI for the assignments they currently design.
AI Literacy as the New Essential Skill
The schools developing the most sophisticated AI policies are treating AI literacy — the ability to use AI tools effectively, critically, and ethically — as the essential outcome that education should provide rather than a prohibited tool that education should protect against. This shift produces a fundamentally different institutional response: instead of ‘how do we stop students from using AI,’ the question becomes ‘how do we ensure students can use AI well, know its limitations, and maintain critical judgment about its outputs.’
AI literacy curriculum covers: understanding what AI can and can’t do (its capabilities and failure modes, covered across multiple articles in this series), evaluating AI output critically rather than accepting it uncritically, prompting effectively to get useful outputs, understanding the data and privacy implications of AI tool use, and the ethical dimensions of AI use (attribution, plagiarism boundaries, appropriate contexts). These are genuinely new skills that traditional curricula don’t cover and that educational institutions are actively developing frameworks to teach.
The Equity Dimension
AI tools are not equally accessible across the educational spectrum, and the access gap creates new dimensions of educational inequality if AI provides meaningful educational advantage. Students with reliable home internet, personal devices, and subscriptions to AI tools (many of the most capable tools have free tiers but limited free access) are better positioned to leverage AI for their education than students without these resources. Schools that actively integrate AI tools with institutional access address this gap; those that passively respond to individual student AI use without addressing access may inadvertently amplify existing inequalities.
The specific equity concern that’s attracted the most attention: if AI tutoring provides significant educational benefit (more practice, more immediate feedback, more personalised explanation), and if access to AI tutoring is correlated with socioeconomic status, then AI tools could amplify the educational advantage that already correlates with family income. Institutions that take this seriously are actively providing AI tool access through school accounts, training teachers to use AI for differentiated instruction that benefits all students, and ensuring that AI literacy education reaches all students regardless of personal device access.
