Is AI cheating? Wrong question.

I want to tell you about a brain I know really well. I’ve sat across from it, in a hundred variations, for more than twenty years.

An ADHD brain holds a lot. Big ideas. Whole systems. Real insight about how things work and how they could work better. Connections firing all day long, in the shower, in the car, at 11pm when the house finally goes quiet.

And between all of that and the page? A bottleneck the size of a straw.

Watch what happens when that person sits down to write. The idea that felt whole an hour ago scatters into forty pieces. They start organizing the pieces and three new ideas show up demanding attention. They lose the thread. Find it. Lose it again. Some of their best thinking lives and dies inside their own head. Not because it isn’t good. Because getting it out and into order costs more than the brain has to spend that day.

Now watch what changes when the right support shows up. They talk. They dump every thought, out of order, half finished, contradicting themselves, all of it. And AI holds the pieces still while they arrange them. It organizes. It sequences. It trims. The thinking is theirs. Every idea, every judgment call, every “no, that’s not what I meant, try again.” Writers have had editors doing this exact work for as long as there have been writers, and nobody has ever called an author a cheat because an editor tightened the manuscript. What used to stay locked in one head is now out in the world where people can actually use it.

So. Was that cheating?

Because that’s the argument everyone is having right now. Parents asking whether the essay still counts. Teachers feeding papers into unreliable detectors. Employers squinting at cover letters. Coaches watching clients hand their whole planning process to a chatbot and calling it growth.

Almost all of it stuck on one question: is AI cheating?

That question goes nowhere. The honest answer is “it depends,” and nobody can tell you what it depends on. So the argument loops. One side yells “it’s a tool.” The other yells “it’s a shortcut.” Both are right. Neither is helpful.

Here’s the question that gets somewhere:

Is AI building this person’s independence, or borrowing it?

That question comes straight out of the CSO Framework™, the model I’ve built over two decades of working with neurodivergent teens, young adults, and families. CSO stands for Capacity, Skill, and Ownership. It exists to answer one thing: is this support right for this person, in this moment, for this goal? I built it long before anyone was arguing about chatbots, which matters. This isn’t a test invented to excuse a tool. It’s a test the tool has to pass. And as you’ll see, plenty of AI use fails it.

But first we have to fix the mistake almost everyone is making.

---

Glasses and training wheels are not the same thing

Nobody asks whether glasses are cheating. Nobody says “you’ll never learn to see on your own if you keep wearing those.” Glasses are an access support. They make participation possible. They are not supposed to fade, and we don’t judge them by whether the person eventually goes without.

Training wheels are different. They exist to come off. They’re a development support. We judge them by whether the person is learning to ride, and if they’re still bolted on five years later, something went sideways.

AI can be either one.

For that brain, AI organizing a flood of thoughts is glasses. The ideas were already there. The judgment was already there. What was missing was access to their own thinking in a form other people could receive. For someone else, AI writing the whole assignment is training wheels that never come off. Same tool. Completely different function.

Most of the AI panic comes from collapsing those two things into one category. We accuse people of cheating when they’re actually accommodating. Or we celebrate productivity when someone is quietly outsourcing the exact skill they were supposed to be building. You can’t evaluate the support until you know which kind it is.

So. The test.

---

Capacity: can they access the task?

Capacity is the door. Before anyone can build a skill or own an outcome, they have to be able to get in the room.

When AI works at the Capacity level, it reduces barriers. It quiets the overwhelm enough to start. It breaks a giant task into pieces a tired brain can hold. It gets the thoughts out of the head and onto the table where they can finally be worked with.

This is the glasses lane. If AI is genuinely providing access, it doesn’t have to fade. A person who always uses AI to organize their thinking before a hard conversation isn’t failing at independence. They’re using an accommodation. That’s allowed. That’s the point.

---

Skill: is competence growing?

This is the training wheels lane, and it has a simple test.

Could they do it tomorrow without AI?

If AI explained, modeled, gave feedback, and the person can now do more than they could last week, learning happened. AI supported skill.

If AI produced the thing while the person watched, no learning happened. AI replaced skill. The work got done and the person got weaker. That’s not support. That’s a really polite theft.

The tomorrow test only applies when the goal is learning. That’s why the fork matters. Asking “could they do it without AI” about an access support is like asking someone to earn their glasses.

---

Ownership: who is actually in charge?

This is the layer almost every AI conversation skips, and it’s the one I care about most.

Who decided the goal? Who looked at what AI produced and judged whether it was any good? Who chose what to keep and what to cut? Could the person explain why they accepted the suggestion?

Go back to that brain. Ask the person why a line in the finished piece stayed and they can tell you what it’s doing, what they cut before it, and why it earned its place. The direction, the standards, the final call. Theirs. That’s ownership intact. AI worked for them.

If AI is setting the direction, generating the options, and making the choices while the person clicks accept, ownership is shrinking. And a person who doesn’t own their work doesn’t grow from it, no matter how good the output looks.

And yes, the line blurs. AI suggests a phrase, you absorb it, and a week later it feels like yours. That blur is exactly why ownership is a practice, not a permission slip. You keep checking who’s driving, because the answer can change in the middle of a task.

Ownership applies in both lanes. Always. Glasses or training wheels, the human stays the decision‑maker.

---

The two-minute test

Run this on any AI use, any person, any task:

1. Which lane is this? Access support (making participation possible) or development support (building a skill)?
2. Is it doing its lane’s job? Access lane: can they participate now in a way they couldn’t before? Development lane: could they do more without AI tomorrow than they could yesterday?
3. Who’s driving? Did the person set the goal, judge the output, and make the final calls? Could they defend their choices?

Lane named, job getting done, human at the wheel: the support is appropriate. Anything else: adjust the support. Not the person. The support.

---

Three honest footnotes

The lane call is a conversation, not a self-report. Every kid on earth will claim glasses. Deciding what’s access and what’s avoidance takes an honest look at the person, the goal, and the pattern over time. Parents, teachers, and coaches belong in that conversation.

Existing rules still apply. If a classroom or workplace prohibits AI for a task, this test doesn’t override that. What it does is give you a better conversation than “gotcha.” It’s the difference between policing a tool and developing a person.

Some things we rightly hand off forever. Adults delegate taxes, oil changes, and contracts without shame, and no one calls it failed independence. The same is true with AI. It all depends on the goal. When a student hands an essay to AI, the point was never the essay. It was learning to think and write, so delegating it defeats the purpose. When a professional hands a routine report to AI, the report was never the point either. It was standing between her and her actual work. Same act. Opposite meaning.

---

The whole thing in one sentence

Appropriate support increases access today or capability tomorrow, and always leaves the person in charge.

If you’re worried that someone you love is getting weaker behind a screen, you’re paying attention. Don’t waste that attention on the cheating argument. Run the test. Ask which lane the support is in, whether it’s doing its job, and who’s driving.

Some brains are full of things the world needs and short on ways to get them out. When support meets a brain like that in the right lane, with the person still holding the wheel, it isn’t cheating.

It’s the door finally opening.