Schools Are Teaching Students to Question AI, Not Just Use It

By Suad Seferi · Aug 25, 2026

The first reaction to ChatGPT in many schools was to block it. Now, some educators are trying something more difficult: teaching students not to automatically believe it. Schools across the United States are beginning to treat AI literacy as a basic classroom skill, with lessons focused not only on how generative AI works, but also on where it gets things wrong, how to verify its answers and when it should not be used at all. Recent reporting by the Associated Press found that educators are increasingly moving beyond the early debate over whether students should be allowed to use ChatGPT. The question now is how young people can use these systems without handing over the thinking itself. In Charleston County, South Carolina, teachers attending AI training this summer were shown an AI-generated map of the world. At first glance, it looked convincing. A closer look revealed misspelled countries, places in the wrong locations and names that did not exist. The point of the exercise was simple: AI can produce something that looks authoritative while being wrong. That lesson is becoming more important as chatbots become part of how students search for information, write, study and solve problems. AI literacy is more than knowing how to prompt Charleston County School District, which serves around 50,000 students, spent the past year developing an AI policy with input from teachers, students and parents, according to the AP. The district is now moving into training for teachers and students, including lessons on checking AI-generated information, recognizing fabricated material and understanding what happens to information shared with a chatbot. Similar efforts are appearing elsewhere. According to the AP, 37 US states have issued some form of guidance on artificial intelligence in education. Utah has gone further by appointing an AI education specialist and training thousands of teachers. The direction is significant because much of the early conversation around AI education focused on learning how to use the tools. But knowing how to write a good prompt does not necessarily make someone AI literate. A student can know how to generate an essay, summarize a chapter or ask a chatbot for research and still have little understanding of whether the answer is accurate. AI literacy increasingly means knowing that a chatbot can invent a source, reproduce bias, misunderstand a question or present uncertain information with complete confidence. It also means knowing when using AI is the wrong choice. Some schools are taking a cautious approach Chicago Public Schools provides an example of how complicated that balance has become. The district says students currently do not have unrestricted direct access to generative AI at school, citing concerns around safety, privacy and equitable use. But that does not mean AI has simply been removed from the classroom. Chicago's guidance starts with AI literacy rather than tool use. Teachers can demonstrate generative AI, examine its outputs with students and use it in supervised learning activities. The district also tells students that AI should support learning rather than replace independent thinking. When approved AI tools are used for academic work, students are expected to identify AI-generated material and explain how they used it. The district separately warns that approval of a tool for privacy and security does not mean its answers can automatically be trusted for accuracy, bias or fairness. That distinction may become increasingly important. A chatbot can be technically safe to access while still giving a student a bad answer. Teaching children to doubt the machine For schools, the challenge is no longer simply whether students will use AI. Many already do, and they will encounter more of it outside the classroom regardless of school policy. The harder task is teaching them what to do when an answer appears instantly on a screen and sounds completely certain. That includes checking sources, comparing information, recognising when something does not make sense and being comfortable rejecting the answer altogether. These are familiar critical-thinking skills. Generative AI has simply made them more urgent. The same conversation is relevant well beyond the United States. Schools across the Balkans are also dealing with students who already have access to ChatGPT, Gemini and other AI tools, often before teachers or institutions have developed common rules for their use. The region does not necessarily need to begin with more AI tools in classrooms. It may need to begin with a simpler question: do students understand why the machine in front of them can be useful and still be wrong? Teaching someone how to ask AI for an answer is easy. Teaching them when to question that answer may be the more important skill. Associated Press: “Schools are starting to teach AI literacy. For many, that means helping kids see chatbots’ flaws,” published August 21, 2026. Chicago Public Schools: Student Access to Generative AI and AI Guidance for Students.