Students Use AI Every Day. Their Teachers Were Never Trained.

Ninety percent of college students now use AI in their coursework. That number reads like progress until you sit with the rest of the study, because the same research found that nearly half of the educators standing in front of those students have never had a single hour of formal AI training.¹ Students sprinted. Preparation stayed at the starting line. That distance has a name, and naming it is the first step to closing it: the Readiness Gap.

The Readiness Gap is not a story about lazy teachers or reckless students. It is a story about sequence. We handed the tool to everyone and scheduled the training for never. The predictable result is a classroom where the person expected to guide the technology knows less about it than the eighteen-year-old in the third row.

The numbers are not a scandal. They are a to-do list.

Here is what the 2026 research actually shows. Roughly two-thirds of educators already use AI in their work, so this is not a question of resistance.¹ The problem is that adoption ran ahead of instruction. Only about one in ten teachers has completed comprehensive AI training, and a large share have completed none at all.¹ When you ask educators to rate their own confidence with these tools, the average lands near the middle of a five-point scale, roughly 2.5 out of 5.² That is not incompetence. That is what it looks like when capable people are asked to lead in a subject nobody taught them.

Now, you might be thinking that this closes on its own as people tinker. It does not. Casual tinkering builds habits, not judgment, and judgment is the entire job. A teacher who has quietly figured out how to draft an email with a chatbot has not thereby learned how to spot a confident, fluent, completely wrong answer in a student’s lab report. Those are different skills, and only one of them shows up on a syllabus.

The gap everyone actually agrees on

The most useful finding in the study is the one that reads like common sense: students, parents, and teachers already agree on where AI belongs. They accept it for support work such as explaining a hard concept, brainstorming, or finding resources. They reject it for the things that require human judgment, like grading and final academic decisions.¹ Two out of three educators and students also share the same worry, that AI sounds confident while being wrong.²

Sit with that, because it flips the usual framing. We keep treating AI in education as a fight to be refereed. The data says the boundary is already drawn in people’s heads. What is missing is not agreement. What is missing is the practice that turns a shared instinct into a repeatable classroom habit.

The boundary is not the hard part. Everyone already knows where it goes. The hard part is practicing it until it becomes second nature.

The Harvest Kernel take

Why one big training day will not fix this

The instinct, when a gap this size appears, is to schedule a mandatory professional development day and call it handled. It will not be handled. A single workshop in August is a fact dump, and facts about a fast-moving tool go stale before the leaves turn. Worse, a one-time session teaches the tool of the moment instead of the underlying skill of working with any tool, which is the only thing that keeps up.

This is where the Readiness Gap gets closed the same way it should have been built in the first place, through SeedStacking: small, deliberate, repeatable practice that compounds. Not a curriculum overhaul. One honest rep at a time. A teacher who runs the same assignment through an AI tool once a week, watches where it helps and where it lies, and names that line for students is building durable fluency. Do that for a semester and you have something no workshop delivers, which is calibrated instinct.

If you want proof that structured, human-led practice beats a firehose of access, look at what happened when 1,600 teachers were given AI mentors rather than just AI logins. Access was never the missing piece. Guided reps were. The same lesson is playing out at the policy level, where California moved to make AI literacy a civic foundation rather than an optional elective.

What a prepared educator actually does

Preparation here is smaller and more concrete than most policy documents admit. A prepared educator does four things, and none of them require a computer science degree. They know how these systems fail, so a confident wrong answer does not slip past them. They can reverse-engineer an output enough to ask where it came from. They set an explicit line for each assignment between help and outsourcing. And they judge a tool by whether it deepens understanding, not by how fast it finishes the task.

Notice that every one of those is a teaching move, not a tech skill. That is the reframe that dissolves most AI anxiety in the room. You are not being asked to become an engineer. You are being asked to do the thing you already do, which is guide judgment, in one new domain. The question of who steers the classroom when AI walks in is not new either, and educators keep landing on the same answer: the teacher still holds the pen.

Close your own gap first

If you are an educator reading this, the honest starting move is not to fix the system. It is to close your own Readiness Gap by a few inches this week. Run one assignment through an AI tool before your students do. Watch what it gets right, catch what it gets wrong, and write down the line you would draw. That is the whole first rep. The confidence you are missing is not a personality trait you lack. It is reps you have not done yet, and reps are the most fixable thing in education.

The students already showed us they will not wait. The kind thing, and the professional thing, is to catch up on purpose rather than by accident. Small seeds, planted on schedule, are how a gap this wide gets crossed.

The mistake that quietly widens the gap

There is one move that makes the Readiness Gap worse, and it is the most tempting one: ban the tool and hope the problem waits. It will not wait. A ban does not remove AI from the room. It removes the teacher from the conversation. Students keep using it anyway, just without guidance, and the educator forfeits the one thing that makes them valuable here, which is a seat at the table while judgment is still forming. Prohibition feels like control. In practice it is abdication, and it hands the most important literacy of the decade to whoever the student happens to trust instead.

The alternative is not a free-for-all. It is presence. Stay in the room, set the line for each task, and model the judgment out loud so students can watch a skilled adult decide. That is harder than a ban and far more useful, and it is exactly the kind of small, repeatable move that SeedStacking is built to make routine.

Sources

1. Instructure, “The State of AI in Education 2026” research (via PR Newswire), July 2026.
2. Higher Ed Dive, “90% of students use AI in the classroom, Instructure poll finds,” July 2026.