AI Cheating Isn’t the Problem (These 3 Classroom Habits Are)

AI cheating is real, it’s growing, and schools everywhere are grasping in the dark, trying to find solutions where they don’t exist.

Let’s look to the numbers and see where we stand:

• Educators worry - In a recent Carnegie survey, 78% of educators said they worry about AI cheating, with 33% very worried, and 45% somewhat worried.
• Students adopt AI - A 2025 College Board survey found 84% of U.S. high school students are already using generative AI for schoolwork.
• No one’s guided - Rand findings show that 81% of those students aren’t getting guidance on how to work with AI from their teacher.
• Cheating emerges - Pew Research Centre cites that 59% of U.S. teens say students at their school are using AI chatbots to cheat.

Most students are now using AI for school. Almost none of them are being guided on how to use it. And AI cheating is coming through on the other side.

Traditional schools have come up with a predictable response: detection tools, stricter AI policies (for the 22% of schools that have them…), prohibition enforcement that students - and AI - have already outpaced.

But what if AI cheating was never the problem we thought it was?

If you’re trying to figure out how to stop AI cheating in your classroom, your school is likely pointing you in the wrong direction. We’re here to share why cheating shouldn’t be your target, and the 3 classroom habits you should be focusing on.

AI Cheating Is the Wrong Diagnosis

Schools keep treating AI cheating as THE emergency. But it’s more of a diagnostic for an emergency that existed long before AI.

When a student uses AI to produce an answer that doesn't look like AI, most people flag AI as the problem. But the sharper diagnosis is that the assessment was measuring the wrong part of the learning process.

SparkNotes, essay mills, friends who wrote the paper, they all existed long before AI hit the scene. And they all did effectively the same thing, just at a lower efficiency.

Output - preferred by industrial education - is, and has always been, a brittle proxy for understanding.

AI didn’t introduce the flaw - it just arrived with better tools and made the flaw harder to ignore.

Schools that respond to AI cheating with detection tools and prohibition policies are fighting a burst pipe with a wet mop. The intention is there, but the method is wildly misaligned with the solution.

Focusing on cheating will NOT solve AI cheating. You need to fix the problem at the source.

Habit 1: Measuring Outputs Instead of Thinking

TL;DR: Output-based grading is an enabler of AI cheating. Process-based assessments are the structural fix.

The dominant assessment model in K-12 is a straightforward one: produce a thing, get graded on that thing.

Essay, project, test - the logic holds across almost every format. The output is the measure, and the thinking that produced it is assumed, not tested.

This classroom habit weaselled its way through the industrial era, teaching students, school year after school year, that their job is to manufacture an acceptable output.

Production > Reasoning

We set that focus, then students did what students do and learned to cognitively optimise.

Before AI, that meant memorizing the cookie-cutter process, drafting to minimum word counts, mimicking the surface features of analytical writing. Heck, the marking guides we share are literally a reverse-engineered map for getting from A to B most efficiently!

The only real change with AI is that the optimization layer has become mostly trivial.

If you want to derail AI cheating, make thinking the outcome you measure.

Process-based assessments are the structural fix that makes that possible. Forcing deep learning by pushing the spotlight onto questions like:

• What was your reasoning at each stage?
• What did you consider and reject, and why?
• Where did you get stuck - and how did you push through it?
• Where and how was AI used to solve problems?
• If you used AI as a sparring partner, what did it get wrong?

By focusing on the process over the output, students naturally perform the reps of deep thinking. And when this happens, AI becomes an enabler, not an attack vector for dishonesty.

Habit 2: Training Passive Reception Instead of Active Thinking

TL;DR: Passive instruction builds passive thinkers. Passive thinkers become passive AI users.

The constraints of traditional education have forced two distinct roles within the classroom.

Role 1: Teachers as the primary source of information.

Role 2: Students as passive consumers of that information.

Let me be clear, this is an entirely understandable consequence of the overloaded classroom. One teacher cannot reasonably provide individualized attention to thirty students simultaneously.

The only logical decision is to teach everyone at once by aiming at the mythical ‘middle.’

But students trained through the lecture-based model are inevitably rewarded for quiet absorption over questioning and reasoning. And the cognitive muscles that matter most - the ability to sit with difficulty, push against an idea, and arrive at understanding through genuine effort - never get exercised.

When students trained to be passive get their hands on AI, they just continue the relationship.

Ask AI → Get Answer

There’s no reason for a student to work collaboratively with AI when they’ve been trained to accept information without question. This is how we end up with the AI-answer-engine.

The only practical solution is to train active engagement.

And the research around this is unambiguous.

A 2023 meta-analysis covering studies across K-12 found that active learning carries an effect size of 1.005 on academic achievement and 1.204 on learning retention. For context, strategies yielding an effect size of 0.04 (the ‘hinge point’ for effectiveness) and above are usually considered to be ‘highly effective’.

If you want to stop passive AI cheating, you have to start creating an environment that blends active participation with AI literacy.

Habit 3: Designing Against Collaboration

TL;DR: Schools train isolation, and isolation left students with no model for AI collaboration.

MacKenzie Price - Co-Founder of Alpha School & 2 Hour Learning - points a finger directly at the operating principle behind most traditional classrooms:

The embedded message is unmistakable: working with others outside of a controlled environment is a violation of the rules.

This habit carries compounding costs in the AI era.

Number 1: It systematically suppresses the skills that are most distinctly human and most durable - leadership, storytelling, relationship-building, and the ability to work through disagreement toward a shared goal.

These are not skills AI is acquiring. The student who develops them has a compounding advantage that the student trained to work alone (in silence) does not.

Number 2: It leaves students with no model for what productive collaboration looks like - because they’re forced out of it.

The productive AI user operates the way a great student operates in a real study group. They bring their own thinking into the room, they push back on and adjust to what surfaces, they leave sharper than they arrived.

Human interaction trains AI collaboration.

If you're going to design collaboration out of the classroom, don't act confused when students misuse the most powerful collaborative thinking tool ever built.

Fix the Habits, and AI Cheating Dissolves

The student cheating problems schools are fighting will keep coming back as long as the habits that created them stay in place.

Each of our three habits builds a condition that AI cheating depends on.

❌ You measure output. AI can produce the output. AI cheating works.

❌ You train passive reception. Students reach for AI passively. They default to the easy answer.

❌ You design against collaboration. Students have no model for AI as a thinking partner. They use it as a ghostwriter instead.

But don't make the mistake of thinking any of this is new.

All three of these habits were moving in the same direction long before AI arrived. AI is just smart enough to make exploiting them a lot easier.

But this doesn't have to be the end of the story, and these same three decades old habits can also be levers for something different.

Teachers willing to adjust course and change their classroom culture - redesigning assessments around process, building classrooms that demand active thinking, and treating collaboration as a skill rather than a violation - will find that AI cheating largely takes care of itself.

Not because AI use disappears, but because fixing these three habits makes misuse the less optimized path.

AI didn’t introduce a new problem. It arrived in a classroom optimised for the wrong outcome. Smart teachers build something different.

Check out the education roles on offer through Crossover.