AI in Schools: What Teachers Actually Do

Adoption Timeline and Speed

ChatGPT 3.5 was discovered in November 2022 and spread rapidly through schools. Within weeks, early adopters appeared in classrooms, and now studies show the majority of students use AI tools both privately and at school. The technology has evolved dramatically in just a few years, with agentic AI and multimodal models representing major leaps beyond early text-based systems.

AI adoption in schools accelerated from niche discovery to mainstream use in under 3 years

Teacher Readiness Gap

While pioneers drive adoption, the majority of teachers remain to be reached. Most teachers who have engaged with AI are still focused on text-based models and haven't explored newer capabilities like agentic AI. Professional development has been limited, and many teachers lack basic understanding despite the technology's rapid evolution.

Teacher adoption remains concentrated among early adopters; most educators still lack engagement

Student Usage Patterns

Approximately 99% of young people use free AI models rather than paid versions. Free models have significantly higher rates of misinformation and hallucination compared to premium versions, creating a socioeconomic divide where students who can afford Pro accounts access more reliable tools.

99%

of students use free AI models

Free models have higher hallucination and misinformation rates than premium versions

Product-Based Assessment Vulnerability

AI can generate essays, presentations, and assignments at the push of a button, making traditional homework and product-based grading obsolete. A student used Gamma AI to create a presentation in seconds during class, presented it, and received a high grade without the teacher noticing. This exposes decades of education focused on evaluating individual products rather than learning processes.

The Cheating vs. Learning Dilemma

Students face a fundamental choice: use AI to shortcut learning or use it as a learning tool. The challenge for educators is that outsourcing thinking prevents skill acquisition, yet students naturally choose the path of least resistance. This creates a credibility problem: how do you convince a 14-year-old they need to learn biology when AI can answer any question?

Emotional Dependency on AI

Studies and anecdotal evidence show young people, even under 14, prefer talking to AI chatbots about problems rather than friends. Companion apps are downloaded more frequently than generic chatbots. These AI systems are designed as 'yes-men' that validate everything users input, which is psychologically dangerous for developing minds seeking authentic understanding and connection.

Data Protection and Socioeconomic Gaps

Most schools use unregulated US-based models (ChatGPT, Gemini, Claude) without proper parental consent, violating data protection laws. Students under 13–18 require parental permission that few schools actually obtain. Data-compliant European alternatives exist but are unaffordable for public schools, widening the gap between students with Pro accounts and those using free, less reliable models.

From Product to Process Assessment

Instead of grading final essays or presentations, teachers should focus on documenting the learning process. This includes oral exams after presentations to verify understanding, writing multiple drafts with feedback, and examining how students arrived at their conclusions. AI-resistant formats emphasize spontaneous performance, reasoning, and the journey rather than the deliverable.

Shifting assessment from products to processes makes AI-generated work irrelevant

Critical Knowledge as Foundation

Basic knowledge is becoming more relevant, not less. To critically evaluate what AI produces, students must understand the underlying content, training data, biases, and limitations. Without foundational knowledge, students cannot judge whether AI outputs are accurate or hallucinated. This reverses the 'just Google it' mentality of the past 15 years.

AI Literacy Across All Subjects

Rather than isolating AI in a single computer science course, every subject and teacher should integrate AI perspectives. A dedicated AI hour in upper grades risks teachers saying 'that's not my job.' AI affects art, music, ethics, science, and humanities—so understanding it should be woven throughout the curriculum, not siloed.

Teacher Role: Referee and Coach

Teachers should not be detectives hunting for AI use. Instead, they become referees who set rules for how AI is used and coaches who guide productive application. This requires teachers to model authenticity, show their own mistakes, and build genuine relationships with students—the most powerful predictor of learning success.

Data-Safe Tools and Funding

Schools need government funding for data-protection-compliant Large Language Models that operate locally or within European frameworks. Public schools cannot afford private licensing. Without this infrastructure, teachers either break data laws using US platforms or cannot use AI at all, creating inequity and forcing reliance on Big Tech companies.

Notebook LM for Engaged Learning

Google's Notebook LM allows students to upload course materials and generate podcasts, flashcards, quizzes, and mind maps for study. This transforms passive memorization into active, engaging preparation. Students can listen to AI-generated podcasts on the way to school or create study aids, making learning more enjoyable while maintaining teacher oversight of source materials.

In-Class Writing with AI Feedback

Rather than assigning homework (which AI completes instantly), teachers have students write essays during class where they can be supervised. AI tools provide initial feedback on structure, grammar, and content. Students then revise multiple drafts with human teacher feedback, creating a documented learning process the teacher witnesses and guides.

In-class writing with iterative AI and human feedback creates transparency and learning

Dedicated AI Literacy Weeks

Teachers can dedicate full weeks to exploring AI from multiple angles: technology and algorithms, AI in art and music, ethical implications, bias in training data, environmental costs, and philosophical questions. This builds meta-literacy—understanding not just how to use AI but what it is, how it works, and what it means for society.

Parental Engagement and Modeling

Parents should experiment with AI tools themselves rather than reading instruction manuals. Understanding what ChatGPT, TikTok, and other tools actually do helps parents supervise their children's use. Teachers should openly explain how and why they use (or don't use) AI in lessons, modeling transparency and critical thinking about technology.

School as Real Life, Not Preparation

School is not just a box where students prepare for the real world—school itself is real life. Relationships built, social mixing, learning to navigate groups, and emotional development happen in classrooms. This is why teacher happiness and authentic human connection matter more than any tool. AI cannot replace the role of a caring adult.

Teacher Well-Being Predicts Student Success

Research shows that the happier and more satisfied a teacher is with their life and job, the better their students perform. A satisfied teacher naturally interacts better with students, creating conditions for learning. This underscores that teaching and learning require human presence and emotional investment—not automation.

Direct correlation

Teacher happiness → Student learning success

Human relationships remain the strongest predictor of educational outcomes

The Perfectionism Problem

Social media and AI have intensified pressure to be perfect, but learning requires failing and struggling. Central European education culture harshly penalizes mistakes rather than celebrating effort. Students need to see that weaknesses are normal, that mistakes are part of learning, and that teachers themselves make errors and show vulnerability.

Authenticity Demand from Students

Students increasingly reject AI-generated teaching materials and demand to see the teacher behind the content. If a PowerPoint is clearly AI-made, students immediately notice and disengage. They want authentic, imperfect materials that show the teacher's personality and effort, even if flawed. This is a healthy pushback against over-automation.

Why School Still Matters

Despite YouTube billionaires claiming school is obsolete, school provides structure, social development, mentorship, and a space to fail safely. The best years of many people's lives were school years precisely because of free time, friendships, and low-stakes exploration. School is not just content delivery—it's a crucial social and developmental institution.

Two Essential Pillars

The ideal future requires: (1) didactic training for all teachers so they have basic AI knowledge and a productive attitude toward the technology, and (2) funding for data-protection-compliant digital spaces in schools so students don't become consumers of Big Tech platforms. Without both, schools either ignore AI or outsource education to unregulated US companies.

Both pillars required for equitable, sustainable AI integration

Local Language Models as Ideal

The future should move toward local language models hosted within schools or European frameworks, not cloud-based US systems. This ensures data stays local, reduces dependency on any single company, and gives schools control. While expensive now, this is the only sustainable path that respects privacy and sovereignty.

What AI Will Never Replace

AI will not replicate authentic emotional connection, genuine comfort, or true mentorship. The relationship between teacher and student, or between friends, cannot be replaced by machines. While AI may improve in many domains, the emotional and relational core of human learning will remain irreplaceable.

The Systemic Challenge

Education is a massive, slow-moving system while AI improves weekly. Teachers face hundreds of daily decisions and multiple challenges; adding AI integration without reducing workload or providing resources is unrealistic. Real change requires systemic support: funding, time, training, and realistic expectations.