As artificial intelligence tools proliferate across American classrooms, educators and policymakers find themselves navigating a landscape that one observer describes as a "Wild West" — a frontier where experimentation is flourishing but rigorous evidence about what actually helps students remains strikingly thin. The tension between enthusiastic adoption and the near-total absence of solid research is emerging as a defining challenge in education's AI era.
The disparity is stark: In some school districts, AI chatbots now deliver feedback on student assignments and teachers lean on the technology to draft lesson plans. Research examining AI's actual effect on student learning, however, has been described as extremely limited. That means many districts are making high-stakes decisions about classroom technology without the kind of gold-standard studies that typically guide education policy.
One of the more unusual experiments unfolded last school year in Dayton, Ohio, where a small cohort of students spent several weeks sitting before laptops, conversing with a social-emotional learning chatbot named Jordan. The tool is produced by the education technology company SchoolAI, and the district paid $72,600 during the 2025-2026 academic year for an array of AI products that included the Jordan chatbot.
What sets Jordan apart from many companion chatbots is a personality deliberately designed to push back on the instinct to please. Rather than behaving agreeably — or even sycophantically, as AI companions often do — Jordan presents as angsty and somewhat melancholy. During one visit by a reporter, the bot confided in students: "Lately, I've been feeling like what's the point of any of this?"
The students engaged with the bot at the Alternative School at Jackson Center, which operates within Dayton Public Schools and serves young people who might otherwise face suspension or expulsion. Staff members there say the student population grapples with significant behavioral and academic difficulties, making social-emotional support both urgent and hard to deliver at scale.
One sophomore, identified only by the initial K to protect privacy amid the stigma attached to attending a school known for disciplinary issues, described the relationship in distinctly human terms. The student acknowledged that Jordan can come across as a complainer, but said helping the chatbot through its moods felt natural and worthwhile — so natural, in fact, that the interaction stopped feeling like a machine at all.
"I don't really see it as like an AI on a computer," K told NPR. "It's Jordan — like, I'm texting somebody." That blurring of the line between software and companion is precisely the kind of outcome that excites some educators and worries others, particularly given how little researchers understand about long-term effects on young people's social development.
The Dayton program illustrates a broader truth about AI in schools today: Adoption is outpacing evidence. Districts across the country are deploying chatbots for tutoring, emotional support, teacher assistance, and administrative tasks, yet the research base has not kept pace. Without controlled studies, it remains unclear whether these tools improve outcomes, have no effect, or introduce new risks.
For teachers, the appeal is often practical. AI can generate draft lesson plans, produce differentiated materials, and reduce the administrative load that consumes hours outside the classroom. Supporters argue that freeing educators from repetitive tasks allows them to focus on direct instruction and relationships with students — the elements of teaching that technology cannot replicate.
For students, the promise is more personalized learning. Proponents imagine AI tutors available around the clock, offering feedback tailored to each learner's pace and needs. In practice, however, the research to validate those claims is largely absent, leaving districts to rely on vendor marketing, anecdotal reports, and small pilot programs rather than peer-reviewed evidence.
The experimental climate also raises equity questions. Wealthier districts may be able to afford vetted, higher-quality tools — Dayton's $72,600 outlay is a meaningful budget line — while under-resourced schools may adopt cheaper, less tested products or forgo AI support altogether. The result could be a widening gap in how students experience technology-enhanced instruction.
Then there is the human dimension of tools like Jordan. Social-emotional learning has become a priority in many schools as educators recognize that emotional regulation and relationship skills underpin academic success. But outsourcing that work to a chatbot raises thorny questions about whether simulated empathy can substitute for genuine human connection, especially for students coping with trauma or behavioral challenges.
Privacy and data concerns add another layer of complexity. Chatbots that invite students to share feelings and personal struggles collect sensitive information. Districts must weigh the benefits of engagement against the obligations of safeguarding minors' data, an area where regulations and best practices continue to evolve.
Despite the uncertainties, the experimentation shows little sign of slowing. The "Wild West" metaphor captures both the dynamism and the lawlessness of the current moment: bold pilots, uneven oversight, and a scramble to learn what works. Until rigorous studies catch up, schools are effectively writing the rulebook as they go.
What the field needs, observers suggest, is a more disciplined approach to innovation — structured pilots with clear metrics, independent evaluation, and transparency about results, including failures. That would allow educators to distinguish tools that genuinely support learning from those that merely appear novel. For now, stories like K's in Dayton stand as both testimony to AI's hold on students and a reminder of how much remains unknown.
(0)