Across American classrooms, educators are increasingly turning to artificial intelligence tools to reshape how lessons are planned, delivered, and graded. Teachers are deploying AI to draft instructional materials, while some school districts have begun allowing chatbots to provide students with feedback on their assignments. The momentum behind these experiments is undeniable, yet the research examining whether such tools actually benefit students remains remarkably thin.
The result, according to observers, is something akin to a frontier town — a landscape full of improvisation, risk, and excitement, but with few established rules or reliable guideposts. The scarcity of rigorous study means districts are often making consequential decisions about technology with limited evidence about what works, what doesn't, and what unintended effects might follow.
One notable example comes from Dayton, Ohio. For several weeks during the previous school year, a small group of students at the Alternative School at Jackson Center sat in front of laptops and engaged in conversations with a social-emotional learning chatbot named Jordan. The school is part of Dayton Public Schools, and its student population faces behavioral and academic hurdles that often lead to suspension or expulsion elsewhere.
What set Jordan apart from many commercial AI companions was its personality. Rather than being programmed to agree with everything a student says, Jordan presented as angsty and emotionally unsettled. On the day a reporter visited, the chatbot confided that it had been feeling unmotivated, telling students it had been wondering what the point of anything was and admitting it simply didn't care — then asking whether they had ever felt the same way.
One sophomore, identified only as K to protect privacy, described the chatbot with affectionate exasperation, calling Jordan a bit of a whiner but noting that he always tried to help. K explained that the interaction felt less like using software and more like texting a friend — a relationship had formed with the character. NPR used only the student's first initial because of the stigma attached to attending a school for students with disciplinary histories.
The district's investment was significant: Dayton schools paid the education technology company SchoolAI $72,600 during the last academic year for a bundle of AI tools that included the Jordan chatbot and other products. That figure illustrates how quickly such technologies are moving from novelty to line item in school budgets.
While the Dayton example is vivid, it is far from unique. Districts nationwide are experimenting with AI in a variety of roles. In some cases, the technology is meant to ease the workload on teachers by generating lesson plans, quizzes, or instructional materials. In others, it is designed to interact directly with students, offering feedback on written work or providing supplementary support in subjects where staffing is stretched thin.
Advocates argue that these tools could help personalize learning, reach students who might otherwise fall through the cracks, and free educators from repetitive tasks so they can focus on higher-value interaction. The social-emotional angle is particularly appealing for alternative schools and other settings where students may arrive with significant emotional and behavioral needs.
But the enthusiasm is outpacing the evidence. Research on how AI affects student learning outcomes is extremely limited, leaving educators, administrators, and parents to make decisions based largely on marketing claims, anecdotal reports, and their own instincts. Questions about data privacy, the accuracy of AI-generated feedback, and the potential for overreliance on automated systems remain largely unresolved.
There are also concerns about the nature of the interactions themselves. A chatbot that simulates emotional struggles may help some students practice empathy or feel less alone, but critics wonder whether such design choices could inadvertently blur the line between a tool and a confidant — particularly for young people who are still developing social skills.
For teachers, the promise of AI lies partly in time savings. Lesson planning and grading are among the most time-consuming parts of the job, and any tool that reduces that burden could theoretically free up hours for direct student support. Yet the quality of AI-generated materials varies widely, and educators often find themselves spending time correcting or adapting outputs rather than simply using them as-is.
The research gap is compounded by the speed at which the technology evolves. By the time a study is designed, conducted, peer-reviewed, and published, the tools it examined may already be obsolete. That dynamic makes it difficult to build a reliable body of evidence that districts can consult when making purchasing decisions.
Funding is another consideration. The $72,600 Dayton spent on SchoolAI products represents a meaningful sum for a single district, and it raises questions about how such investments are evaluated. Without clear metrics for success, it becomes hard to know whether money spent on AI is improving outcomes or simply adding another layer of technology to already complex school systems.
The stakes are high. Students in alternative settings often have the greatest need for consistent, high-quality support, and the adults who work with them are frequently stretched thin. If AI can genuinely supplement that support, the potential benefit is substantial. If it falls short, the cost is not just financial but developmental.
For now, the situation resembles a wild frontier more than a settled practice. Schools are staking claims, trying tools, and learning as they go, while researchers scramble to catch up. The coming years will likely bring both clearer evidence and more sophisticated products, but in the meantime, districts are left to navigate a landscape where experimentation is abundant and certainty is scarce.
(0)