How Teachers Really Feel About AI in Schools

The One Thing Teachers Want You to Know About AI

A few months ago, I called a middle school teacher in Ohio to ask about her experience with ChatGPT. She laughed. "I've used it to write a parent newsletter," she said. "And I've caught three kids using it on their book reports. That's about it."

That conversation stuck with me because it didn't match the headlines. The news was full of stories about AI "transforming education" or "destroying academic integrity." But teachers, the people actually in classrooms, seemed to have a more complicated, quieter relationship with these tools.

Now there's data to back that up. A 2023 study by Thomas K. F. Chiu, a researcher at The Chinese University of Hong Kong, surveyed 88 school teachers and leaders about how generative AI tools like ChatGPT and Midjourney were actually changing their work (Chiu, 2023). The results are not what you'd expect from the breathless coverage. They are more useful, more human, and more honest.

Chiu found that teachers are not panicking. They are not celebrating either. They are doing something far more interesting: They are trying to figure out what their job even means anymore.

The Four Domains Where AI Actually Hits

Chiu's study began with a framework from a systematic literature review on AI in education, which identified four domains where AI matters: learning, teaching, assessment, and administration. That sounds academic, but it is actually a useful map. Each domain represents a different kind of pressure point.

The teachers and leaders in the study filled out a survey and then joined focus groups. They talked about ChatGPT for text generation and Midjourney for image generation. They were not theoretical. They had used these tools, or their students had.

What emerged were four themes, each with three subthemes. But the real story is not the taxonomy. It is what the teachers said about their own roles.

Learning: Students Are Already Ahead

The first theme was about how students learn with AI. Teachers reported that students were using ChatGPT to generate explanations of concepts they did not understand. Some used it to summarize readings. Others used it to brainstorm ideas for projects.

This sounds like a win. But here is the twist: Teachers said students often did not know when the AI was wrong. They accepted its confident, fluent answers as truth. One teacher in the focus group described a student who submitted a ChatGPT generated essay about a historical event that never happened. The AI had hallucinated a battle. The student had no idea.

Chiu's analysis found that teachers saw a "new prerequisite knowledge" emerging: Students need to learn how to evaluate AI outputs critically. This is not just about fact checking. It is about understanding the difference between a tool that predicts the next word and a tool that knows the truth.

Teaching: The Know It All Attitude

The second theme was about teaching itself. This is where things get uncomfortable.

Teachers reported that the presence of AI changed how they prepared lessons. Some used ChatGPT to generate quiz questions or to rewrite explanations for different reading levels. Others used Midjourney to create visual aids for history or science classes.

But the most striking finding was what Chiu called the "know it all attitude." Teachers said students sometimes challenged their expertise by citing ChatGPT's answers. A student might say, "But ChatGPT says the correct answer is..." and the teacher had to decide whether to trust their own knowledge or the machine's output.

This is not a technical problem. It is an authority problem. When an AI can generate a plausible answer to almost any question, the teacher's role shifts from being the source of knowledge to being the validator of knowledge. That is a hard shift to make, especially for veteran teachers who spent decades building their expertise.

Assessment: The End of the Take Home Essay

The third theme was assessment, and it was the most practical. Teachers said traditional homework, especially essays and written assignments, became nearly impossible to evaluate fairly. They could not tell if a student wrote the essay or if ChatGPT did.

Some teachers tried to fight this by requiring in class writing. Others changed their assignments to focus on process rather than product. One teacher described asking students to submit their ChatGPT prompts along with their final essays, then evaluating how well the student had guided the AI.

Chiu's study suggests that assessment is the domain where AI will force the most immediate policy changes. Schools cannot simply ban the tools. Students will use them anyway. The question is how to design assessments that measure what students actually know and can do, not what an AI can generate for them.

Administration: The Hidden Time Saver

The fourth theme was administration, and it was the most surprising. Teachers reported that AI saved them significant time on tasks like writing emails, drafting parent newsletters, and creating lesson plans. One teacher said they used ChatGPT to write a draft of an individualized education plan for a special needs student, then edited it to fit the student's specific situation.

This is the part of the story that gets lost in the panic about cheating. Teachers are overworked. They spend hours on paperwork that has nothing to do with teaching. If AI can reduce that burden, it might actually make teaching more sustainable.

But Chiu also found a catch. Teachers worried that using AI for administrative tasks would erode their professional judgment. If a teacher relies on ChatGPT to write a parent email, do they lose the skill of communicating with parents? If an AI generates a lesson plan, does the teacher understand why that plan works?

What Teachers Actually Want From Policy

The study did not just describe problems. It offered three specific suggestions for practice and three implications for policy. These are worth reading carefully, because they come from teachers, not from technologists.

Practice: Three Things Teachers Need

First, teachers need to adopt a "know it all attitude" about AI. That does not mean pretending to know everything. It means being willing to learn alongside students and to admit when they do not know something. The teacher who says "I don't know, let's ask ChatGPT together and then check its answer" is modeling the skill students actually need.

Second, teachers need to teach "new prerequisite knowledge." Students need to learn how AI works at a basic level. They need to understand that it is a statistical language model, not a mind. They need to practice evaluating its outputs for accuracy, bias, and completeness.

Third, teachers need to embrace interdisciplinary teaching. AI does not respect subject boundaries. A student using ChatGPT for a history paper might need to understand something about computer science, ethics, and writing. Teachers from different subjects need to coordinate their approaches.

Policy: Three Things Schools Need to Change

First, assessment policies need to be rewritten. The old rules about plagiarism and academic integrity assume that students produce their own work. That assumption is no longer valid. Schools need clear policies about when and how students can use AI, and those policies need to be enforced consistently.

Second, schools need to offer AI education as part of the curriculum. This is not just about teaching students to use ChatGPT. It is about teaching them to think critically about the technology. Chiu's study suggests that AI literacy should be treated as a core competency, like reading and math.

Third, schools need to update professional standards for teachers. The current standards assume that teachers are the primary source of knowledge in the classroom. That assumption is no longer true. Professional standards should include competencies around AI, such as knowing how to evaluate AI generated content and how to design assessments that work in an AI rich environment.

What This Research Does Not Prove

Chiu's study is qualitative. It is based on 88 participants, all from one educational context. That means the findings are not statistically generalizable to all teachers everywhere. The teachers who volunteered for this study were likely more interested in AI than the average teacher. The study captures their experiences and perspectives, but it does not measure the prevalence of those experiences across a larger population.

The study also focused on ChatGPT and Midjourney specifically. Those tools are changing rapidly. What was true in 2023 may not be true in 2025. The findings about assessment, authority, and professional identity are likely to persist, but the specific technical details will shift.

The study did not measure student outcomes. It did not test whether using AI improves learning or harms it. The teachers reported their perceptions, but those perceptions may not match objective measures of student achievement.

These are not flaws in the study. They are honest limits. The value of Chiu's work is not in providing final answers. It is in giving us a clear, grounded picture of how teachers are experiencing a technology that is changing their profession whether they like it or not.

The Open Questions That Matter

The study raises several questions that deserve more research.

How do teachers' attitudes toward AI change over time? The teachers in this study were early adopters. What happens when AI becomes mandatory, when school districts require its use?

How does AI affect teacher retention? If AI reduces administrative burden, it might keep teachers in the profession longer. If it undermines their authority or forces them to constantly adapt, it might drive them away faster.

How do students actually learn with AI? The teachers in the study observed students using AI, but they did not measure learning outcomes. Does using ChatGPT to explain a concept improve understanding, or does it create a false sense of comprehension?

What happens to students who do not have access to AI? The digital divide is real. If some students use AI to write essays and others do not, the gap in academic performance could widen.

These questions are not academic. They will determine how AI shapes education for the next generation.

What This Actually Means

Here is what I take away from Chiu's study, as a journalist who has read hundreds of papers about technology and education.

• ▸Teachers are not the problem. The narrative that teachers are resisting AI because they are technophobes or Luddites is wrong. The teachers in this study were using AI, thinking about it, and trying to adapt. Their concerns were professional, not personal. They worried about their students' learning and their own integrity.

• ▸The real fight is about authority, not cheating. The cheating panic is a distraction. The deeper issue is that AI challenges the teacher's role as the expert in the room. That is a harder problem to solve than catching students who copy and paste.

• ▸Assessment needs to be rebuilt from scratch. Multiple choice tests and take home essays are both vulnerable to AI. The only assessments that will survive are those that measure process, collaboration, and real time performance. That is a massive shift for a system built on written exams.

• ▸AI might make teaching more sustainable, but only if we let it. The administrative time savings are real. Teachers could use AI to reduce their workload. But they need permission and training to do that. They also need to trust that using AI does not make them less professional.

• ▸The most important skill is not using AI. It is evaluating AI. Students need to learn how to check an AI's claims, how to recognize its biases, and how to decide when to trust it and when to doubt it. That skill is not currently taught in most schools. It needs to be.

The teachers in Chiu's study were not excited about AI. They were not terrified either. They were doing what good teachers always do: They were trying to figure out how to help their students learn in a world that had suddenly changed. That is not a headline. It is the quiet, difficult work of education.