ChatGPT Can Write the Essay, But Can Your Student Defend It?

Every professor knows. Every teacher knows. Most of them just don't want to say it out loud.

Written assignments are broken.

Not because students are lazy. Not because education is failing. But because artificial intelligence has made it virtually impossible to tell the difference between a student's original work and something generated in thirty seconds by a machine.

And the tools designed to catch AI-generated writing? They don't work reliably. They flag innocent students. They miss sophisticated users. They create an arms race between detection algorithms and generation algorithms that detection will never win. It's like trying to catch counterfeit money when the counterfeiter has a better printer than the mint.

So here's the question that every educator needs to ask right now: if you can't trust what a student writes, what CAN you trust?

The Uncomfortable Reality

Let's stop pretending this is a future problem. It's happening right now, in every institution, at every level.

A 2024 survey of university students found that over 50% had used AI tools on assignments. Not to cheat, necessarily. Many used it to brainstorm, outline, or "clean up" their writing. But the line between "assisted" and "generated" is blurry, and it gets blurrier every semester as the tools get better.

Professors are responding in predictable ways. Some have banned AI tools entirely, which is about as enforceable as banning calculators in 1985. Some have switched to in-class handwritten exams, which tests penmanship more than thinking. Some have added oral defense components, which is the right instinct but often amounts to five minutes of softball questions at the end of term.

None of these solutions address the real problem. The real problem isn't AI. The real problem is that written assignments were always a proxy for thinking, and now the proxy is broken.

An essay was never the point. The point was to demonstrate that a student could research a topic, organize their thoughts, construct an argument, anticipate counterarguments, and communicate clearly. The essay was just the vehicle. But now that AI can produce a perfectly serviceable essay on any topic in any style at any reading level, the vehicle is useless. The question is whether we can assess the underlying skills directly, without the proxy.

The answer is yes. But it requires a fundamental shift in how we think about assessment.

What Can't Be Faked

Here's what ChatGPT cannot do. It cannot stand in front of a room and defend an argument in real time against a challenger who is actively trying to poke holes in it.

Think about what that requires. The student has to actually understand their position, not just have words on a page. They have to anticipate objections before they're raised. They have to think on their feet when an unexpected challenge comes. They have to structure their response in seconds, not hours. They have to maintain composure when their logic is questioned. They have to listen carefully to what the challenger is actually saying, not just wait for their turn to talk.

No AI tool can do this for them. There's no earpiece, no hidden screen, no way to phone a friend. It's just them, their knowledge, and their ability to think clearly under pressure.

This is what oral assessment looks like when it's done right. Not a casual Q&A after a presentation. Not a rehearsed defense of a thesis statement. A structured, rigorous, real-time evaluation where the student's thinking is tested live, under conditions that make faking impossible.

Why Most Schools Don't Do This

If oral assessment is so effective, why isn't everyone doing it?

Three reasons.

Scale. A professor with 200 students cannot individually assess each one in a live debate format. The time required is prohibitive. Even with TAs, the logistics are overwhelming.

Consistency. Human evaluators are subjective. Two different professors watching the same student performance might give wildly different scores. There's no standardized rubric that can account for the nuance of a live argument.

Training. Most students have never been taught how to argue. Not in the formal sense. They don't know how to structure a rebuttal. They don't know the difference between attacking an argument and attacking a person. They don't know how to concede a point gracefully and pivot to a stronger one. Throwing them into a live assessment without training is setting them up to fail.

These are real obstacles. But they're solvable. And AI, ironically, is the solution.

AI as the Practice Partner

Here's where things get interesting. The same technology that broke written assessment can fix oral assessment.

Imagine a student preparing for a debate on climate policy. Instead of writing an essay that ChatGPT could write for them, they open a platform and start arguing their position against an AI opponent.

The AI doesn't agree with them. It challenges every claim. "You say carbon taxes reduce emissions, but what about the economic impact on low-income households?" The student responds. The AI follows up. "That's a valid point about rebates, but the data from British Columbia's carbon tax shows mixed results on income equality. How do you reconcile that?"

The student has to think. Actually think. Not copy, not paste, not prompt-engineer their way to a good grade. They have to know their material, organize their thoughts in real time, and defend their position against an opponent that knows the counterarguments.

After the session, they get feedback. Not "good job" or "B+." Specific feedback. "Your opening argument was strong but you abandoned it after the second objection. Your rebuttal on economic impact relied on a single data point. You spoke for 45 seconds without addressing the opponent's core claim." They can see exactly where they need to improve, and they can practice again immediately.

Now multiply this across a semester. A student who practices defending their ideas against an intelligent AI opponent twice a week for four months is going to develop critical thinking skills that no written assignment has ever produced. They'll walk into the final assessment not with a memorized script but with genuine competence.

What This Looks Like in Practice

ELBO's NOVA module was designed for exactly this scenario.

Students select or are assigned a topic. They argue their position against an AI that adapts to their level, pushes back on weak arguments, and introduces real counterevidence. The session is scored across multiple dimensions: the strength of their evidence, the structure of their arguments, their ability to address objections, their clarity of expression, and their composure under pressure.

Professors can assign specific topics, set parameters for difficulty level, and review performance data across their entire class. They can see which students are improving, which are struggling, and exactly where the gaps are. Not based on a paper that might have been written by a machine, but based on a live performance that can only come from the student themselves.

The student's ECHO profile tracks their growth over time. By graduation, they don't just have a transcript full of letter grades. They have a demonstrated record of critical thinking, argumentation, and oral communication skills that any employer or graduate school can verify.

The Shift That's Coming

This isn't a nice-to-have. This is an inevitability.

The institutions that figure out AI-proof assessment first will attract better students, produce better graduates, and build better reputations. The ones that keep relying on written assignments and plagiarism detection software will fall behind, slowly at first, then all at once.

The essay isn't dead. But the essay as the primary tool for assessing student thinking? That's done.

The students who will succeed in the next decade aren't the ones who can write the best paper. They're the ones who can stand up, open their mouth, and defend what they believe with evidence, logic, and clarity. Whether the audience is a professor, a hiring manager, or a room full of skeptics.

Can your students do that today?

If not, it might be time to give them a place to practice.