AI Can Prep Your Case. It Can't Save You at the Lectern

What happens in competitive debate when AI meets live performance

I arrived home exhausted this past Saturday evening after a long day doing what I’ve done almost a dozen times a year for well over a decade - judging, coaching, and/or running a secondary school debate tournament. A personal bonus was that my 22-year-old daughter was there with me. She was a top debater in both middle and high school, learning the activity entirely in a pre- generative-AI world, and now coaches for rival schools in both leagues where we compete. There are few experiences as rewarding as watching your adult child thrive in the same working environment where you can swap stories and talk shop as peers.

Walking through the door, the first thing I saw on the counter was the latest edition of The Economist with a cover story about AI rewiring childhood. The timing felt almost too perfect, because what I’d witnessed all day was another example of how AI is infiltrating an academic activity in which I’ve been involved for over fifteen years.

What’s interesting about discussing AI in the context of competitive debate is how it reframes the “cheating” conversation. Using AI to help prepare for an upcoming tournament isn’t cheating in the traditional sense. No one is submitting AI-generated work to a teacher for a grade. There are no grades - just two teams in front of a judge and possibly some spectators, and about 40 minutes to make your case.

And yet something has shifted - dramatically - in the culture of debate preparation over the past two years. This change was most visible for me at the middle school tournament, where I spent Saturday and where I hadn’t judged a full MS event in over a year. Instead of the wacky variety of MS arguments and exuberant buzz of the “prep” period, I encountered homogenized cases and near-silence as students furiously copied word for word from pre-canned AI debate language onto colored sheets of paper, the only permissible material allowed into the round.

In the debates themselves, many students struggled to pronounce words they had never used and make arguments they didn’t understand. Although they appeared more prepared than ever before - detailed note sheets covered with text - in reality many could not articulate a coherent response in the face of any kind of rebuttal. I’m confident AI is the main culprit.

A Quick Note on Format

For readers unfamiliar with competitive academic debate: there are dozens of formats across middle school, high school, and college - Lincoln-Douglas, Policy, Public Forum, Parliamentary, all with local variations on each.¹ The MS format I’ve coached for fifteen years has features that turn out to be particularly revealing in the AI era.

In our league, students receive topics about a month in advance. They can do as much preparation as they want during that time - research, case-building, writing out arguments. This has always been allowed, long before ChatGPT existed. For our most recent tournament, the topics were NYC’s plan to operate city-run grocery stores, the ethics of excavating historical burial sites, and whether private corporations or government institutions pose a greater threat to free speech. These are tough topics for middle schoolers, but representative of issues we’ve debated in our league since its inception.

Here’s the twist: on tournament day, students don’t know which specific motion they’ll debate, or which side they’ll argue, until twenty minutes before the round. During those twenty minutes, they can transfer notes from their prepared materials onto colored paper - a different color for each round - to prevent reading pre-written speeches from their screens. Whatever they write in those twenty minutes is all they can bring to the lectern.

The format was designed to reward deep preparation while still requiring genuine understanding. You can’t memorize a script because you don’t know your side until the last moment. And critically, because students aren’t permitted access to their computers, they needed to actually understand their arguments well enough to transfer those thoughts onto paper in a way that makes sense when speaking aloud.

This format turns out to be a remarkably effective stress test for AI-assisted preparation.

Pre-Tournament Prep Prior to COVID

Before advanced LLMs, preparation prior to the event meant searching for articles and learning about a topic from the ground up. We would spend practices breaking down the motion - what are the main points of clash? What arguments might you make on either side? Then we would find articles, conduct research through trial and error, and build our knowledge base.

This is how my daughter learned to debate in the 2010’s - engaging in discussions during practice, building arguments from sources she actually read, and reviewing, adjusting, and learning what worked through actual competition. Most significantly, when she and her teammates prepped during the tournament itself, it was lively, dynamic, and interactive.

Arguments develop organically and individually. Middle school students vary in abstract reasoning skills. Some make concrete, simplistic arguments while older 7th and 8th graders are starting to understand nuances around First Amendment protections or how higher taxes fund government services.

The purpose of preparation was to allow students to understand issues at their own pace. The best way to learn was through interactive discussion with a coach - asking questions, practicing arguments, learning new vocabulary - all the kind of scaffolding important for any kind of learning. When they reached the debate room, at least the arguments were their own and they could explain them on their terms.

As a coach and judge, I was always amazed at the sheer variety of arguments students made on the same topics - often marveling at ingenious approaches from other teams. Why didn’t we think of that? This made judging interesting and coaching challenging. My debaters frequently encountered unexpected arguments, which meant better opportunities to think on their feet - another way to separate strong debaters from novices.

When the world went online in 2020, debate followed suit. A debate culture of reading from your screen took hold. And then AI came along.

Frictionless Prep in the Age of AI

In fall 2025, within minutes of opening an LLM, any MS student with even rudimentary prompting skills can generate a sophisticated case with solid arguments, detailed evidence, and clear structure. The AI will anticipate counterarguments and provide rebuttals. Merely dropping the motion into ChatGPT produces a reasonably coherent case for both sides.

While creating cases is exactly what debate preparation is supposed to produce, the process is supposed to take weeks, the result of hard work and earned knowledge. When there’s literally no effort required to generate detailed arguments, all the worst aspects of using AI come into play. No friction in finding sources. No requirement to think through the topic by reading articles. No struggle to turn ideas into clear argumentation.

Gone are the discussions, the novel thinking, the questions that revealed significant gaps in knowledge. AI prep gives students a Formula One racing car when many still need training wheels.

For new debaters, especially younger ones in middle school trying to learn fundamentals, it’s destructive. And because most AI-assisted prep happens at home rather than in practice, coaches have limited visibility into how students are actually preparing. This is a new challenge - one that I imagine debate programs across the country are only beginning to grapple with as the scale of the problem becomes clear.

As I glanced around the cafeteria during prep time, virtually every screen was filled with AI-generated text. Emojis were the dead giveaway, but even from ten feet away I could see dense text with the familiar bullet-point structure that’s a telltale sign of AI work.

In Round-Tournament Prep, Then and Now

In earlier years, many newer MS debaters would be caught off guard during their first tournament prep period. Though coaches constantly emphasize working on their cases prior to the tournament, many students don’t internalize what’s going to happen until they’re in it.

Typically, half the debaters in the room would be frantically realizing they hadn’t prepped nearly enough as they watched more experienced debaters work through cases they themselves had created, urgently passing colored paper back and forth, talking constantly as they explained their cases to each other. The room would be alive and abuzz with low-grade panic as these newer students tried to conjure arguments they’d had weeks to think about. It was a critical lesson.

In my daughter’s time, debate round prep was an incredibly valuable part of the day.

Now? Even the youngest and most inexperienced debaters have mounds of “prep” on their computers - usually massive and dense, structured into exactly the required format (assertion, reasoning, evidence), complete with citations and vocabulary few 6th graders understand. On Saturday, that hum of activity was missing. Every debater was furiously copying text many had probably never actually read, with enough detail to cover both sides of multiple sheets. What previously required more careful deliberation - what to include, what to prune, what to focus on - became a copying exercise. I had never experienced such quiet at a MS debate tournament.

Students entered debate rooms feeling more prepared because they had paper covered with text but, in reality, many had little idea what they’d copied or what they were supposed to be arguing.

How It Translated Into Actual Debates

What I observed across multiple rounds Saturday was striking sameness. Almost all the arguments were identical - the same three points on each side, the same evidence cited, the same structure - because it was all being drawn from the same pool of training data.² Having judged hundreds of rounds over the years, I’m used to encountering a wide range of arguments, including creative approaches that surprise me. That variability was largely absent. Everyone had conformed to some median, and that median was obviously AI-generated. I judged four rounds on two motions and saw schools use largely the same cases for each. The variety and creativity simply weren’t there.

But sameness wasn’t the only problem. The much larger issue was that too many debaters clearly didn’t understand the arguments they were making.

I had watched students frenetically copying material during prep - more material than they possibly could have processed. They didn’t understand words they hadn’t written themselves. They mispronounced terminology they’d never said aloud. When opponents asked clarifying questions during cross-examination, most students simply repeated scripted language or retreated into platitudes revealing they didn’t understand their own arguments or worse, carefully scrutinized their paper as if the answer would magically appear.

One motion involved First Amendment issues. I watched three sixth graders attempt highly sophisticated constitutional arguments (strict scrutiny!) when it was clear they couldn’t pronounce “constitutional” - let alone explain what the First Amendment actually protects.

The Paradox: No One to Cheat But Yourself

I realize I’m being hard on the debaters but the difference in just two years was stark. Reading cases prepared by someone else has been a staple and Achilles heel in debating for decades. AI enables it on a scale we’ve never seen.

But here’s what makes this different from the classroom AI conversation: in debate, there’s no one to cheat but yourself.

When a student submits an AI-generated essay to a teacher, the transgression is clear, even if detection is difficult.

Debate is different. Teams are expected - even encouraged - to collaborate and research together. No teacher is being deceived. No grade is at stake.

But there is a final arbiter: a live audience - a judge, the opposing team, sometimes spectators - watching you perform in real time. AI can write your speech. It cannot deliver it. It cannot answer a point of information from the other team. It cannot pivot when you’re presented with an unexpected argument. It cannot respond in real time as the debate evolves, with different issues rising or falling based on new definitions, rebuttals, or frameworks.

No student speaking in the final two speeches of a debate can possibly use AI to create a coherent synthesis of the debate that actually happened. Plenty try. When third speakers pre-write their final speeches using AI, the disconnect is obvious. Points are emphasized that were barely touched in opening speeches. New arguments - a blanket prohibition in every format - pop up because the AI script has no idea what actually occurred.

The exposure is immediate and public. There’s no delay between performance and feedback. You’re not waiting for a grade. You’re standing at a lectern watching a judge’s face as you read through arguments you never internalized and demonstrate you barely understood the debate.

If you used AI to generate a case you don’t understand, the only person you’ve cheated is yourself. You’ve traded the hard work of learning material for the illusion of preparation. And that trade-off becomes obvious the moment you open your mouth.

The Silver Lining

I don’t want to leave this piece on a note of despair, because what I saw Saturday also gave me genuine hope. Debate is hard. It’s supposed to be challenging. We should not be surprised that 11, 12, and 13 year-olds want as much of a safety net as they can get when speaking in front of a group on a challenging topic. The problem is they don’t realize cutting corners will not lead to improvement.

The students who truly understood their material shone. They rose above the AI-sameness precisely because they could engage with ideas rather than recite them. When the other team read a complex, likely AI-generated argument, they could respond substantively. When asked difficult points of information, they had actual answers. Their arguments were grounded in reasoning they could trace, evidence they could explain, and positions they could defend under pressure because they understood the topic.

Whether these students used AI to help them research is irrelevant. What matters is that they did the important work: they actually read, reviewed, and learned what they’d prepared. AI may have aided the process, but it wasn’t a substitute. Used carefully, AI can turbocharge debate preparation. Unfortunately - because students focus on winning, and because policing an extracurricular activity is nearly impossible - too many debaters are using AI to take shortcuts.

But the silver lining for debate - and perhaps for education more broadly - is that the activity will ultimately reward the right students. The debaters who receive awards at tournaments are unlikely to be the ones who outsourced all their thinking to AI.

Why This Matters Beyond Debate

I’ve been writing about AI and education for nearly a year. Over that time I’ve watched students go from tentatively dipping their toe into the AI pool to swimming laps around most adults responsible for overseeing their learning.

What has been documented repeatedly is that, in the right hands, AI can produce genuinely useful output. In a debate context, LLMs are especially good at immediately breaking down issues into pros and cons. But that output can substitute for true understanding in contexts where no one checks. And the younger the student, the less likely they are to check themselves.

AI has made the preparation phase easier. It can generate arguments, find evidence, structure cases. But it cannot internalize that material for you. It cannot make you understand.

Competitive debate is where someone checks. Immediately and publicly.

The Value of Public Speaking

In any activity where you have to perform your understanding live - debate, oral exams, Socratic seminars, real-world professional presentations - AI’s limitations become immediately visible.

That’s why I think public speaking activities - debate, Model UN, mock trial, speech competitions³ - will become increasingly valuable as AI continues to transform education. Assessment formats that require oral defense of claims and arguments are even more authentic than in class written work. Even if students use AI to help them prepare - which will not only be inevitable but even encouraged in some cases - once the questioning starts, they are on their own.

A student standing at a lectern, fielding questions in real time, revealing whether they actually understand their arguments - that’s something AI cannot fake. It’s what my daughter learned through years of hard work before AI existed. It’s what the strongest debaters I judged Saturday still demonstrate.

The question for debate coaches - and for all of us working with young people in 2025 - is how to convey that over-reliance on AI is a trap. AI-generated cases will not lead to wins against the best teams. Like so many things in life, the harder path will eventually lead to the confidence that comes from actually knowing what you’re talking about.

There are no F’s in competitive debate. Just humiliation when you try to discuss a topic you never understood in the first place. For students willing to do the work, avoiding that result should be motivation enough.

---

Interested in MS and HS Debate Leagues in the NY/NJ/CT Tri-State area?

Check out the links below:

New York Debate League (MS)

East Coast High School Public Debate Program (HS)

---

Connect With Me

Beyond this newsletter, I work directly with schools, educators, and organizations navigating AI integration. Take a look at my website and reach out - I’d love to hear what you’re working on.

1. Parliamentary debate, our preferred HS format, is probably the most AI resistant. Parli is known for its impromptu nature - high school teams receive a motion (for example, This House would eliminate the electoral college) and have anywhere from 15 - 30 minutes to prep their cases using only their own knowledge before entering the debate room. These events are typically run on the honor code. Of course some students violate the rules - as long as students have access to the internet, unless coaches are prepared to put their debaters under video surveillance, it’s impossible to prevent a debater intent on cheating. But the kind of debaters attracted to parliamentary debate tend to be students uninterested in using AI - it pits smart high school students against each other to match wits and showcase logic and argumentation skills, not how well you can use AI to find the precise piece of evidence to make your case. Our HS Parli tournaments since AI arrived have been more lively, more fun, and more interesting for everyone involved and one of the reasons we’re considering adding an impromptu motion to our MS events.

2. Few MS students know how to use Deep Research tools or other more advanced techniques to take advantage of ways AI use could give them a further advantage. No one wants debate to become an AI arms race but I suspect that’s what’s happening in HS formats that rely primarily on evidence and statistics.

3. Another event in which I’ve been actively involved over the past 6 years is the National High School Ethics Bowl which is a unique speaking competition which requires teams to present cases but has a very different ethos than a standard academic debate. This year’s cases contain many AI ethical dilemmas.