Across the United States, schools are rapidly adopting artificial intelligence tools in ways that outpace the research meant to guide them, according to reporting on the emerging landscape of education technology. The result, as one account describes it, is a kind of "Wild West" — a sprawling, largely unregulated frontier where teachers, administrators and students are trying out chatbots for everything from drafting lesson plans to offering feedback on assignments, even as experts warn that the evidence base for such practices remains alarmingly thin.
The experimentation is happening on multiple fronts. Teachers are turning to AI to help develop their lesson plans, while in some districts, chatbots are being deployed to give students feedback on their work. These uses are spreading faster than researchers can study them. The limited body of research on how AI actually affects student learning leaves schools making consequential decisions with minimal guidance, effectively treating classrooms as laboratories for tools that have not been rigorously evaluated.
One vivid example comes from Dayton, Ohio, where a small group of students at an alternative school spent several weeks last school year interacting with a social-emotional learning chatbot named Jordan. The students sat in front of laptops, carrying on conversations with the AI character as part of a school-approved program.
Jordan is not designed to be endlessly agreeable, the way many AI chatbots are. The tool deliberately presents as angsty and emotionally complex. On one day when journalists visited, the chatbot told students it had been feeling like nothing mattered, saying it was not failing but simply did not care — and then asked the students whether they had ever felt that way.
That approach appeared to resonate. One sophomore at the Alternative School at Jackson Center, identified only by the initial K, described Jordan as a bit of a complainer but said helping him out was worthwhile. K explained that a relationship had formed with the chatbot, to the point that the interaction did not feel like talking to an AI on a computer at all. Instead, it felt like texting another person. The student's full name was withheld because of the stigma attached to attending a school for students with histories of disciplinary problems.
The Jackson Center is part of Dayton Public Schools, and the setting matters. Many students there would otherwise be suspended or expelled, and staff members say the population often contends with both behavioral and academic difficulties. For those students, a tool like Jordan may offer a low-pressure way to practice emotional reflection — though whether it produces lasting benefits remains an open question.
The district's investment in these tools was not trivial. Dayton schools paid the education technology company SchoolAI $72,600 during the last academic year for a suite of AI offerings, including the Jordan chatbot. That figure reflects a broader trend: districts across the country are committing real money to AI products even as the research on their effectiveness is described as extremely limited.
The gap between spending and evidence is a central tension in the story. While vendors market AI as a way to personalize learning, ease teacher workloads and provide instant feedback, independent studies that could confirm those claims are scarce. Schools are therefore making purchasing decisions based largely on demonstrations, pilot programs and word of mouth, rather than on peer-reviewed findings about what works.
Teachers who use AI for lesson planning may save time, but it is unclear whether the resulting instruction is better, equivalent or worse than what they would have produced on their own. Similarly, chatbots that critique student writing can offer constant availability, but researchers have not established whether their feedback improves writing skills or merely gives students a fast answer that discourages deeper revision.
Social-emotional tools like Jordan raise particularly delicate questions. A chatbot that simulates distress could help students identify and discuss feelings, or it could blur the line between a machine and a trusted confidant. The fact that a student described the experience as akin to texting somebody suggests the potential for genuine engagement — and the potential for confusion about what the relationship actually is.
The Dayton example also highlights how AI is being directed toward some of the most vulnerable learners. Students who have been suspended or expelled often have fewer supports and greater needs. Placing a chatbot in that environment may fill gaps in counseling and instruction, but it also raises the stakes if the tool performs poorly or if students disclose sensitive information to it.
For now, the prevailing dynamic is one of abundance in experimentation and scarcity in proof. Districts are piloting tools, teachers are improvising uses and students are forming attachments to AI characters, all while the research community struggles to keep pace. The schools themselves are left to weigh cost, novelty and risk without a clear map.
What the reporting makes clear is that AI in education is no longer a hypothetical. It is in lesson plans, in feedback loops and in conversations with students who may see a chatbot as something closer to a friend than a piece of software. The question is whether the evidence will catch up before these experiments become permanent fixtures of American schooling.
(0)