Teachers want AI taught — but not trusted blindly

Teachers are not asking whether artificial intelligence will change education. They are asking how schools can respond without sacrificing the thinking, trust and relationships that make learning work. A new nationally representative NPR/Ipsos poll of 545 US K–12 teachers captures that tension clearly: nearly three in four believe AI will have greater implications for education than computers or the internet.

The findings are not simply pro-AI or anti-AI. Six in ten teachers surveyed have used AI for work, most often to create study guides, quizzes or classroom materials. More than half have used it to plan lessons, while others use it for administration, parent communication and differentiated learning plans. Used carefully, AI can return valuable time to an already stretched profession.

Yet the same teachers see serious risks for students. Fifty-four percent say AI makes it harder for children to develop critical-thinking skills, and 55% describe it mainly as a shortcut for avoiding work. Nearly six in ten believe it is eroding trust between students and teachers. Those concerns have already changed practice: about four in ten now require more work to be completed by hand, and a similar proportion require more work to happen in class.

The most important number may be 78%. That is the share of teachers who say responsible AI use should be taught as part of the school curriculum. In other words, concern is not the same as rejection. Teachers recognise that AI is not going away. They want children to learn how to use it without outsourcing the mental effort education is meant to develop.

One classroom example reported by NPR offers a practical model. A biology teacher asks students to use a chatbot to describe an animal and its habitat, then verify the response against reliable sources. The AI answer is not the finished work; it is material to investigate. Students practise prompting, fact-checking and recognising that a confident answer can still be wrong.

That distinction matters. AI becomes harmful when it acts as an answer machine: paste in the assignment, copy the response and submit work the student cannot explain. It can support learning when it acts as a thinking partner: explain a difficult idea another way, generate practice questions, suggest a starting point or produce a claim that the student must test.

Schools can make that line visible with three simple expectations. First, students should say when and how they used AI. Second, they should verify factual claims with trustworthy sources. Third, they should be able to explain every part of the work in their own words. If a student cannot defend the answer, the tool has replaced learning rather than supported it.

The poll also highlights why a blanket rule will not serve every learner. NPR spoke with a special education teacher whose autistic student used AI support to complete work that might otherwise have remained inaccessible. Thoughtful use can remove barriers, personalise explanations and help a child begin. Fair policy must distinguish supportive access from having a machine do capable students' thinking for them.

Parents have a role too. Instead of asking only, 'Did you use AI?', ask, 'What did you ask it, what did it get wrong and what did you decide for yourself?' Those questions replace suspicion with accountability. They also show children that adults do not need to know everything about the technology to set useful boundaries around it.

The lesson from teachers is not that schools must choose between embracing AI and protecting critical thinking. Protecting critical thinking requires teaching AI use properly. Children need regular, supervised practice in questioning outputs, checking evidence, preserving their own voice and knowing when the best choice is to work without AI at all.