87% of Schools Have AI. Only 1 in 4 Have Any Rules for It.

87% of Schools Have AI. Only 1 in 4 Have Any Rules for It.

Walk into almost any classroom today and AI is there — in the lesson plan the teacher built in 20 minutes instead of three hours, in the essay a student submitted, possibly in the feedback that came back. What's largely missing is any official guidance on how any of that is supposed to work. According to data compiled by Lumichats, 87% of K-12 districts report dealing with AI in some capacity, yet only 23% have adopted a formal policy.

More than half of U.S. states have issued some form of AI guidance for schools, but critics — including analysts at GovTech and the Learning Agency — say that guidance tends to be vague, aspirational, and short on the specific decisions schools actually need to make: What tools are permitted? Who owns student data generated by those tools? What counts as AI-assisted work versus AI-generated cheating? Without answers, individual teachers are filling the gap however they can, producing a classroom-by-classroom patchwork with no consistency and no accountability. The schools that figure this out first will be doing something the majority of American districts have not yet done: made a deliberate choice.

Gobble's Take: "We're working on a policy" is not a policy — ask your school's principal exactly where that work stands and when it lands.

Sources: Lumichats · Governing.com · EdWeek

AI Detectors Flag Innocent Students 17% of the Time. Some States Are Finally Saying: Stop Using Them.

A student turns in an essay she wrote herself. A detection tool flags it as AI-generated. Her teacher opens an academic dishonesty case. That sequence is not hypothetical — it's an emerging pattern documented across schools that have adopted AI detection software. These tools carry a false-positive rate of roughly 17%, meaning nearly 1 in 6 accusations could be aimed at a student who did nothing wrong.

The response from education researchers and some state guidance documents has been pointed: don't use these tools. AI for Education's state guidance tracker shows a growing number of state-level recommendations explicitly warning districts away from AI detectors, citing both their unreliability and the punitive environment they create. Non-native English speakers and students with certain writing styles are flagged at even higher rates. The practical alternative — redesigning assignments so that a ChatGPT-generated response simply doesn't fit the task, or having students explain their reasoning aloud — puts the focus back on learning rather than surveillance.

A system that punishes honest students 17% of the time isn't an integrity tool. It's a liability.

Gobble's Take: Before your child's school uses an AI detector, ask them to show you their appeals process — because the odds say they're going to need one.

Sources: Lumichats · AI for Education · CBS News Chicago

Ed3 Synthesized 25 AI Studies on Teaching. The Picture Is More Complicated Than the Hype.

Ed3 published a synthesis of 25 major research studies on generative AI in K-12 education — national surveys, state-level data, behavioral platform analytics, and qualitative research from early 2023 through late 2025. The project, called Between Promise & Practice, is the first artifact in Ed3's Portrait of a Teacher in the Age of AI initiative. Mike Kentz authored a Substack post laying out the findings he considers most important. The headline finding: teachers are using AI heavily for their own back-end work — lesson planning, assessments, communications — not to put AI in front of students. In Utah's statewide dataset of 7,595 teachers, only 17.3% use AI for student personalization. Only 9.5% have created chatbots for student use.

Stanford's behavioral study — tracking what 9,081 teachers actually did on the SchoolAI platform, not what they self-reported — found that experienced teachers shifted away from student-facing tools over time, gravitating toward teacher productivity features instead. Meanwhile, roughly 50-52% of teachers report no formal AI training, and only 19-51% of schools have AI policies. Adoption has outpaced institutional readiness. Critically, zero studies in the set measure whether teacher AI use actually improves student learning outcomes. The benefits are perceived, not demonstrated.

Gobble's Take: Two years into widespread adoption and not a single study can tell you whether any of this is helping kids learn — that gap is the real story.

Source: Mike Kentz Substack

What K-12 AI Policies Actually Look Like When Districts Do the Work

Two districts offer a concrete picture of what it takes to build a real AI policy. In Tucson, the policy covers students, teachers, and staff — grounded in a single premise: AI should enhance and support work, not replace people. That means no copying and pasting outputs. It means reviewing AI responses for fairness and equity. The policy didn't materialize overnight — it was the product of a two-year process that started with a task force drawing from HR, purchasing, communications, and classroom teachers.

Greenwich, Connecticut took a similar approach. Superintendent Toni Jones had a teacher committee study AI policy from Brown University, Harvard, and other institutions before the board's policy committee shaped the final version. The resulting regulation starts with definitions — distinguishing generative from nongenerative AI — then moves through access and permissions, ethical use, data privacy, proper citation, and academic integrity. Jones wanted it readable: bullet points, clear categories, no assumptions about what staff already know.

Both districts also had to manage fear. Jones found that hands-on exposure changed the conversation. "Once teachers actually get in front of it and learn about it, most of them leave very excited," she said. Tucson used a stoplight framework — red for no use, yellow for limited use under specific conditions, green for teacher discretion. That structure has already guided real purchasing decisions, including approved AI use inside Adobe Express and Canva.

Gobble's Take: A two-year task force and a stoplight system aren't bureaucratic excess — they're what it actually costs to protect kids when the technology moves faster than the rules.

Source: EdTech Magazine