CHARLESTON, S.C. - When an instructor asked an AI chatbot to "create a map of the world" in front of a packed high school auditorium, the result drew gasps and laughter. Mali was spelled "Mail." Egypt appeared as "Sopth." In place of Libya sat a country simply called "Africa." Dozens of other names were misspelled or rendered as gibberish.
The demonstration was part of a summer training session for Charleston County School District teachers, and the point was deliberate. "If you ever want kids not to over trust these tools, try the map demo," said Amanda Bickerstaff, founder and CEO of AI For Education, an organization that helps schools draft AI policies and train teachers and students.
After initially trying to ban AI use, a growing number of U.S. public schools are now encouraging classroom experimentation - partly so students can see the technology's shortcomings, including generative AI's tendency to "hallucinate," or fabricate information. AI literacy has become a buzzword of this back-to-school season as educators try to balance preparing students for an AI-driven future with preventing them from outsourcing their thinking to chatbots.
There is no single definition of what AI literacy means or how it should be taught. Tech giants like OpenAI, Google and Anthropic offer schools training sessions in how to use their tools. But among educators, a consensus is emerging that AI literacy goes beyond writing good prompts. "Good AI literacy," said Rebecca Winthrop, director of the Center for Universal Education at the Brookings Institution, "includes knowing when not to use it."
Charleston builds its own approach
Thirty-seven states have published official AI guidance that schools can use as a blueprint. South Carolina is not one of them, so the state's second-largest district set out to create its own, said Lucas Clamp, deputy superintendent of the 50,000-student Charleston County School District. "It has been a huge endeavor."
A year ago, the district contacted AI For Education, which helped draft guidance for Chicago Public Schools, Houston and dozens of other districts. Phase 1 was laying the foundation: developing an AI policy with input from teachers, students and parents. Student focus groups revealed that AI use was widespread but unguided. Teachers and students said they were navigating the technology on their own and wanted clear rules and instruction.
That tracks with national data. Studies show a majority of teenagers and teachers are using AI for schoolwork, but they're teaching themselves how to use it; very few say they have a technical understanding or have received formal training.
Schools are also eager to avoid the mistakes made with social media, where kids explored online spaces without learning about addictive algorithms, comparison culture and other downsides linked to a youth mental health crisis and dwindling attention spans.
As the school year begins, Phase 2 kicks in: training. Teachers and middle and high schoolers will get a mix of online and in-person instruction on how AI tools work and how to use them effectively. The student course walks through familiar scenarios - like asking a chatbot to help with a research paper - to show that generative AI might deliver wildly inaccurate information, including made-up studies, with unflinching authority. "Always verify any factual information you get from GenAI, especially for your classwork," the course advises.
A section on data privacy cautions against sharing personal information with AI tools because conversations can be stored, used for data training and potentially leaked.
High school teacher Ray Knauer attended the summer training and left with "teachable moments" for his AP Research class. He plans to show examples of AI bias and hallucinations to fuel student discussions. "We're all trying to figure this out right now: What's the right balance so we don't scare students away from AI but don't encourage them to run to it," Knauer said. "We have to teach students: You can't just get an answer and move on. You have to use analytical skills and dig deeper."
For educators seeking structured training resources, this AI for Teachers learning path offers practical guidance on classroom integration and assessment design.
Utah takes a statewide approach
Utah was an early pioneer in taking a statewide approach. In 2024, the board of education named Matt Winters as its AI education specialist, making Utah the first state to create a full-time position overseeing the technology in schools.
Over the past year, Winters led AI training for over 7,000 teachers, almost a third of Utah's public school instructors. He is also helping districts shape AI policies, which they are required by state law to have in place by July 2027.
"There is a clear method from the top in Utah," said Chris Agnew, director of the Generative AI for Education Hub at Stanford University. "When you're tackling big change, it's easier to have coherence if everyone is singing from the same hymn book."
A handful of other states have since replicated Winters' post, including Maine, West Virginia and Georgia. Elsewhere, a patchwork of approaches has emerged, with variance from one district to another. Extensive AI literacy training requires steep investment, raising concerns that kids in under-resourced districts might not get the same attention.
In Utah, training starts with the basics of how AI works to teach an understanding of AI bias, hallucinations, ethical concerns and data privacy needs, Winters said. Teachers can then pursue additional training on specific tools.
"If I had to give advice: Figure out how to make training accessible to everyone," said Emma Moss, who oversees AI at the Canyons School District and led an eight-person team that traveled the state doing over 150 AI literacy trainings. The training was funded by a $500,000 AI literacy grant from a local healthcare provider. "We said, 'We will come to you.' Whether you are in a rural area or urban, it didn't matter."
The state has also played a key role in procuring AI tools. Winters negotiated data privacy agreements along with discounted prices, which has allowed rural and under-resourced districts to get access. "Everyone needs to understand what this technology is and how it works," he said, cautioning against blanket bans. "AI literacy is much larger than saying: 'Don't use AI.' It's about how we can prepare kids for the future."
Changing adult mindsets is also key, said Kristina Yamada, the state Board of Education's digital technology specialist. "I think teachers need to change their mentality about students using AI to cheat," Yamada said. There has always been cheating, she said. It's up to teachers to convince students they can't rely on AI for all the answers.
For younger kids, that means starting conversations early: "People can lie. A program can lie. A robot can lie," Yamada said. "For older students it's different: 'Sure, you can use ChatGPT to write computer code. But when the program stops working, you need to be able to fix it.'"
Why this matters for educators
The shift from banning AI to teaching it changes what classroom instruction looks like. Teachers who understand how AI tools work - including their failure modes - can design assignments that require students to verify outputs, question sources and explain their reasoning. Those who avoid the topic risk losing credibility with students who are already using the tools on their own.
School leaders should also note the resource gap: states with centralized AI guidance and training budgets, like Utah, are making progress faster than districts left to build their own programs. Educators in states without official guidance may need to seek out training independently, whether through district initiatives or programs like those offered by AI for Education, to ensure their students aren't left behind.
Your membership also unlocks:
