Taking a cue from trumpets, competitive debate might be charitably taken as a pastime where you put in compressed air and childhood trauma comes out. Sally Rooney, author of Normal People (and one of TIME’s 100 Most Influential People in 2022), describes college debaters in an essay for the Dublin Review as “just students wearing ill-fitting formal clothes.” In the same essay, she also details how competitive debate, as a sport, is an environment where friendless teenagers are made successful and popular, how through her (quite successful) career as a college debater, she inhabited an intense drive for success, but also an underlying craving for social validation, institutional hierarchy. Later in this essay, she says:
It’s hard to imagine my way back into that mindset now. I don’t think I will ever again want something so meaningless so much.
I do think competitive debate gets a bit of a bad rap. It’s important to have an environment for young people that trains critical thinking as well as how to communicate persuasively. There’s evidence that participation in debate improves English language arts test scores in high schoolers. And, debaters do go on to do other things eventually. Ted Cruz, American senator, was the top-speaker in the 1992 US National Debating Championship. Dan Lahav, CEO of Irregular, was the top ranked speaker in the World Universities Debating Championships in 2018. Asher Spector, co-founder of Flapping Airplanes, was the North American champion in 2019. Evan Hubinger, lead of Alignment Stress-Testing at Anthropic, attended Tournament of Champions in 2014 and 2015. Tejal Patwardhan, who leads OpenAI’s frontier evals team, was a semifinalist in the World Universities Debating Championships in 2020.
Competitive debate.
Argumentatively, we really just want to use the above roll call to say: debate is one among many feeder institutions for frontier AI research and governance (among many prestigious careers). Debate also selects and perhaps develops ambitious, verbally capable people. And of course, it is related to how we think and reason about scalable oversight. So, the advent of AI hollowing out debate’s apprenticeship could have consequences beyond tournament culture, and this is a microculture of broader interest.
I spent around eight years doing competitive debate throughout high school and university. And nowadays, I often find myself wondering about that meaninglessness — it felt a great deal at the time like it was the only thing that mattered. I did learn how to speak better, communicate and organize my ideas better, and it did give a shy, bookish boy many friends and positive life experiences. It was good to travel.
And while we’re reminiscing, I would also put forward the question of if debaters in the current age of AI still get to have that experience? Or if, as in the case of Go, even that last luminous layer of wonder and ambition is being stripped away. To summarize some key claims of the afore-linked essay, I would basically put it as:
Go AI became vastly better than humans at playing go.
Some people cheated. While some of this was for self interest, it seems that many began consulting it out of convenience and curiosity.
Cheaters continued to experience themselves as authors of their own play because they selected among its recommendations.
Humans didn’t really improve from Go outside of memorizing AI moves, which is reflected in humans not doing better past move 60.
So, let us spend some time to recount these gears in greater resolution in competitive debate, which is a much different activity. Debate is interactive, requiring delivery, adaptation, teamwork. Does the same vibe prevail?
Competitive Debate
Competitive debating at its core is a constraint optimisation game. You are given a set amount of prep time and a side you have to defend, regardless of what you actually believe about the matter. During this prep time, you are not allowed to Google or use online aid, and are only allowed to consult your partner or tournament officials for clarificatory and administrative matters. After this prep time, you plod into the room, sheafs of paper with hastily scrawled arguments in hand, and give speeches in alternating fashion in front of a judge. Then, after all the speeches are done and all debaters have exited the room, the judges confer amongst themselves, then call the debaters back in to announce the decision and its justification.
Format of British Parliamentary Debate. Taken from here.
There’s much to be said about the metagame here: debating is an activity judged by humans and while there are rules (such as how long speeches should be, or what burdens are fair to impose on a team), these rules are quite underspecified. An excellent and eloquent judge could, in a close room, give a plausible justification for a wide range of possible calls. Further, topics are human decided (either by vote in US formats or by a group of tournament officials (called the Chief Adjudicators in other formats)) and further still, certain arguments and style can be in vogue, and other times, fall out of fashion. In 2021 and 2022, it was in fashion to speak very quickly with a barrage of hard-to-follow arguments, which I chalk up to the pandemic making digital notetaking the norm. Between 2019 and 2023 it was popular to say that a set of reasons you were providing were “structural,” even if they were not actually structural causes in the academic sense of the word.
How could you even be disempowered by AI in such a sport dripping with human influence?
Well, humans have been wondering this themselves for quite some time. Harish Natarajan, a world finalist in 2016 and European champion in 2012, participated in an exhibition debate against IBM’s Project Debater in 2019 on the topic “We should subsidize preschool.” Unlike the well-known successes of DeepMind’s narrow AI against human experts, this was a loss for the AI — Natarajan, in calm and controlled demeanor, persuaded more of the audience to change sides than Project Debater did. Support for Project Debater’s side fell from 79% to 62%, while support for Natarajan’s side rose from 13% to 30%. Although a majority of the audience voted the AI over Natarajan as providing more enriching content (58% to 20%).
And to my eyes as a debater and a judge, he ran circles around the AI — he definitely gave less facts, but was able to frame them more convincingly, scaffold them in ways that were richly structured, and more directly argued for his side of the motion than the AI did.
Harish and the machine.
But the story nowadays is quite different. Hackenburg et al. 2026 found that in dyadic text conversations on UK politics, modern (for the time) AIs were much more persuasive than elite debaters. And these debaters were really good, comprising 56 elite debaters, including four world champions and eleven continental champions, being incentivized financially to be persuasive. But they did also find that models performed comparably to coached elite humans when constrained to roughly human level message length and speed, which gives us a more mixed picture (and this mostly depends on what you consider to be “fair comparison”).
Sidenote: you may observe that this plot seems to quite vindicate elite debaters, as being better than laypeople and professional canvassers.
This is a nice conversation through history! In 2019, specialised AI could marshal information but not persuade as effectively as an elite debater. In 2026, general-purpose frontier AI can outperform elite debaters in text persuasion given equal time, but not when given equal speed and message length. So, overall, the evidence for capability is mixed, but trending significantly in favor of AIs as they get more capable, especially as we see more effort poured into not only coding and math capabilities, but also conceptual reasoning.
But what about the community? What do debaters themselves think, feel, and do? We can look at examples from debate events over the years. This is taking some cues from Debating Seriousposting, an international Facebook group for competitive debaters.
Early on there was some interest in using GPT-3 to generate essay outlines for debate motions. The author estimated that only about 30% were reasonable and relevant, but the rest were either obviously incoherent, answered a different topic, or simply restated the motion without adding anything new. This exercise was repeated with ChatGPT, which was much better but still rudimentary, with repetitive and generic content, but clearly relevant, and reactive to both complicated infoslides and material provided during a round. Although, there was already some concern: “What is alarming is the sheer volume of content it is able to produce from minimal effort which also mitigates a lot of strategic considerations in debating.” and some worry about cheating.
Then, in preparation for the World Universities Debating Championships, a debater describes her use of an AI chatbot to prepare a casefile. She and her partner selected questions and topics, to which the model supplied arguments, examples, intuition pumps, and the resulting casefile was “completely written by an AI chatbot.” During a round about a Pacific Alliance common currency, she mentions reading some of the generated material word for word. Here, some cognition is being deferred, but overall the author reports that the AI casefile is “just like any other casefile”: the debater still has to select when and how to use the evidence in the specific debate that is happening. And in the example, the debater still had to read the whole casefile carefully before the competition, checking it for mistakes.
Several things have happened since then:
DebateData, which I myself have used, launches as a central hub containing motions, videos, tournament information and AI-generated arguments.
Someone proposes generating cases and speech notes for every WUDC video.
Someone proposes a custom “Judging GPT,” trained on one person’s tracking and calls from around forty rounds plus the WUDC scale and EUDC manual, is offered as free feedback.
All the while, AI researchers figure out chain-of-thought, then RLVR, and generally scale up model size and whatnot. Novel progress is made in mathematics, and much of conventional software engineering is successfully automated. By this point, language models have gotten much better, Fable 5, GPT-5.6 Sol, now highly competent reasoners and writers.
Throughout this time, I was also debating, coaching, judging, running tournaments. In debates, I didn’t notice anything out of the ordinary, but in the course of coaching and judging, I would find that some debaters (especially children) would recite obvious AI phrases verbatim, which was a clear red flag. Some debaters would be caught using their phone or laptop in prep time (which was still illegal under tournament rules even before AI), but using it for ChatGPT instead of Google. And as chief adjudicator, many would write their applications for subsidy entirely in AI, which was obvious.
And now, there are governance issues. Major international tournaments have previously collected phones during the elimination rounds to mitigate cheating. But these are later, more prestigious rounds, and the majority of debaters will only be able to do the preliminary rounds, for which it is a logistical nightmare to collect phones. This year, the Australasian regional tournament and the Asian regional tournament in the British parliamentary format implemented a full ban on the use of electronic devices for any purposes during all rounds, and the world championships will be doing the same. To support this, tournament guidance encourages judges to photograph suspected unauthorised tech use, which IMO is quite a powerful image of AI changing the activity even where nobody successfully uses it.
In general, there are recommendations to return to purely paper note-taking and adjudication (to prevent judges from simply transcribing and feeding in a whole round to an AI, and then regurgitating the call), supervised, in-person judge tests, oral interviews about the test rounds, collecting phones throughout all rounds. And maybe giving up on online tournaments entirely, or having separate cyborg and non-cyborg tournaments.
What Now?
You can imagine a world where someone uses AI to: generate every practice case, give feedback after every speech, build AI-generated casefiles, study AI-generated notes from past finals, arrive at Worlds, surrender their phone, and give a completely unaided speech to terrific applause. This doesn’t sound that bad, and this person may have worked quite hard. But I am curious about what goes on in their mind; do they feel empowered, or disempowered?
On one hand, we use technology all the time, such as using the internet to gather data and facts. But we have to resist the urge to be too galaxy-brained here: some crucial part of debating lives in the reasoning, the coming up of arguments. If debate was scored purely on the performance of a speech and the impression it left on the viewer, it would lose something essential that distinguishes it from pure performance arts. Also, some part of it lives in the culture of hanging out with your partner to do a casebuild, to grab a beer after a tournament and not be worried if the guy on the other side is cheating. And some part of it lives in whether or not the debater in question retained valuable critical thinking skills from this process.
In general, how much of the felt value of a sport comes from the (relatively artificial) obstacles? Suits defines a game as the voluntary attempt to overcome unnecessary obstacles. Nobody actually needs to summit Everest without oxygen, or put a basketball into a hoop. But, the pointlessness itself is integral to making the achievement legible. It’s possible that inhaling xenon gas really does increase red blood cell count and improve cardiovascular endurance. In the limit, there is no clear difference between certain drugs and certain tools. A stimulant to help you focus (such as coffee) and a model that provides crucial facts both substitute some amount of capability. Choosing to give 15 minutes of prep rather than 10 is also arbitrary, some formats such as Asian parliamentary do 30 minutes. As such, while there may be differences in degree and character, the line is ultimately down to what the community feels it can tolerate.
Besides, what the debater feels isn’t exactly the right question. The whole point of gradual disempowerment is that the person is making choices that feel locally correct, but are nonetheless objectively disempowered and dependence-cultivating. We could ask it in three different ways:
Are debaters becoming better at unaided reasoning?
It’s not clear, because no one has written this paper yet. We have heuristic points for both sides, like AI giving interactive explanations, and simplifying inaccessible material for ESL speakers, but it could also be that novices never acquire the judgement needed to distinguish good generated cases from slop (there’s some evidence that relying on chatbots to help answer questions reduces persistence and performance in answering questions unaided).
Does the debater still retain agency?
It’s not clear, for conceptually fuzzy reasons. As debaters love to say, there is a spectrum. From AI retrieving a fact for you to AI puppeteering and adjudicating the debate, there’s some line in the middle that’s bad to cross.
Debaters have always used material they did not invent, like coaching, casefiles, conversations.
But AI may also erode critical aspects that are clearly human, like curation, performance.
Is debate still producing the social and cultural goods we value?
Maybe it’s less fun! Maybe tournaments adopt confiscation and surveillance in a way that feels alienating, or maybe it becomes a new normal and feels fine.
Maybe it’s cheaper to prepare, which keeps busy and under-resourced people in the activity, or maybe results reflect access to better, more expensive models.
In the end, it’s not clear either. Debate is still producing debaters, certainly, and people are still attending tournaments. It seems like this is reasonable proof that people continue to enjoy this activity.
In light of lacking or mixed conclusions for each of these questions, what should we conclude from the evidence? Well, some things stand out.
Firstly, it’s clear that the stakes in competitive debate are not the same as something like the end of mathematics. While it has some clear grounding in broader significance to liberal democracy, it is treated by its participants as a sport or a game, rather than an almost spiritual pursuit or a way to build a long lasting legacy. And it’s also not at the stage Go is at: AI has not decisively mastered full competitive debating, or has dominated humans. It is similar to Go in that, from what we can tell, people can still retain a strong experience of authorship after outsourcing generation, at least at the level of case filing.
Further, major tournaments actively protect unaided human performance, and debate remains socially and competitively alive. There is no good evidence yet of population-level skill decline, but we could chalk this up to disempowerment not being at full steam yet.
However, we can claim that substantial parts of preparation and feedback are being automated. In particular, the cost of outsourcing argument generation has collapsed. AI has likely exacerbated cheating at competitions, and such institutions are already incurring significant costs to preserve human-only competition.
Finally, the rate of progress has been extraordinarily fast, and highly in favor of AI. A bespoke, specially trained AI seven years ago could give mediocre speeches, now freely available general models can persuade over text better than elite debaters when unconstrained.
Debating is still alive and healthy, as far as I can tell. Tournament attendances haven’t precipitously declined – in the last world championship, we got quite a large crop of judges and debaters to come to Sofia, Bulgaria.
A gaggle of tournament participants at Sofia WUDC 2026. Evidence of vitality.
To rebut doomerism, people will still travel far and wide to sit in badly ventilated rooms and argue about useless things. I don’t think we have to preserve every inconvenience of debate for its own sake, easier access to information is good, cheap feedback for novice debaters in far-flung communities is good. But the writing is on the wall: it’s getting harder and more costly to stop AI cheating, and many aspects that we like about debate are getting outsourced. There is a future looming, of a debate hollowed out and without much substance remaining.
Provisionally, if I had to give a verdict, I’d say:
Debate has not yet been disempowered. The evidence doesn’t seem to support such a conclusion yet.
But, as AIs get better, we could expect the motivations and feelings of empowerment around debate to significantly change. We would be right to worry that some crucial aspect of debating gets lost if trends continue as they have for the last few years.
Perhaps Rooney is right that debate is meaningless. Certainly, most of the motions are forgotten the next morning, between trophies gathering dust and hangovers blearily clearing. But wanting it so much was not meaningless to me: the wanting was what made a shy, bookish boy learn to speak, think, travel. My worry is therefore not really that future debaters will stop wanting to win, but that winning may cease to require them to become the kinds of people who could have done it themselves.
Thank you to Sam Huang whose initial draft I took some evidence and structure from. Thank you to George Wang, Matt Caito, Dan Murfet, Marina Koljic, Nora Flynn-McIver, Stuart Johnson, and Jordyn Gibson for helpful feedback and discussion.
Taking a cue from trumpets, competitive debate might be charitably taken as a pastime where you put in compressed air and childhood trauma comes out. Sally Rooney, author of Normal People (and one of TIME’s 100 Most Influential People in 2022), describes college debaters in an essay for the Dublin Review as “just students wearing ill-fitting formal clothes.” In the same essay, she also details how competitive debate, as a sport, is an environment where friendless teenagers are made successful and popular, how through her (quite successful) career as a college debater, she inhabited an intense drive for success, but also an underlying craving for social validation, institutional hierarchy. Later in this essay, she says:
I do think competitive debate gets a bit of a bad rap. It’s important to have an environment for young people that trains critical thinking as well as how to communicate persuasively. There’s evidence that participation in debate improves English language arts test scores in high schoolers. And, debaters do go on to do other things eventually. Ted Cruz, American senator, was the top-speaker in the 1992 US National Debating Championship. Dan Lahav, CEO of Irregular, was the top ranked speaker in the World Universities Debating Championships in 2018. Asher Spector, co-founder of Flapping Airplanes, was the North American champion in 2019. Evan Hubinger, lead of Alignment Stress-Testing at Anthropic, attended Tournament of Champions in 2014 and 2015. Tejal Patwardhan, who leads OpenAI’s frontier evals team, was a semifinalist in the World Universities Debating Championships in 2020.
Competitive debate.
Argumentatively, we really just want to use the above roll call to say: debate is one among many feeder institutions for frontier AI research and governance (among many prestigious careers). Debate also selects and perhaps develops ambitious, verbally capable people. And of course, it is related to how we think and reason about scalable oversight. So, the advent of AI hollowing out debate’s apprenticeship could have consequences beyond tournament culture, and this is a microculture of broader interest.
I spent around eight years doing competitive debate throughout high school and university. And nowadays, I often find myself wondering about that meaninglessness — it felt a great deal at the time like it was the only thing that mattered. I did learn how to speak better, communicate and organize my ideas better, and it did give a shy, bookish boy many friends and positive life experiences. It was good to travel.
And while we’re reminiscing, I would also put forward the question of if debaters in the current age of AI still get to have that experience? Or if, as in the case of Go, even that last luminous layer of wonder and ambition is being stripped away. To summarize some key claims of the afore-linked essay, I would basically put it as:
So, let us spend some time to recount these gears in greater resolution in competitive debate, which is a much different activity. Debate is interactive, requiring delivery, adaptation, teamwork. Does the same vibe prevail?
Competitive Debate
Competitive debating at its core is a constraint optimisation game. You are given a set amount of prep time and a side you have to defend, regardless of what you actually believe about the matter. During this prep time, you are not allowed to Google or use online aid, and are only allowed to consult your partner or tournament officials for clarificatory and administrative matters. After this prep time, you plod into the room, sheafs of paper with hastily scrawled arguments in hand, and give speeches in alternating fashion in front of a judge. Then, after all the speeches are done and all debaters have exited the room, the judges confer amongst themselves, then call the debaters back in to announce the decision and its justification.
Format of British Parliamentary Debate. Taken from here.
There’s much to be said about the metagame here: debating is an activity judged by humans and while there are rules (such as how long speeches should be, or what burdens are fair to impose on a team), these rules are quite underspecified. An excellent and eloquent judge could, in a close room, give a plausible justification for a wide range of possible calls. Further, topics are human decided (either by vote in US formats or by a group of tournament officials (called the Chief Adjudicators in other formats)) and further still, certain arguments and style can be in vogue, and other times, fall out of fashion. In 2021 and 2022, it was in fashion to speak very quickly with a barrage of hard-to-follow arguments, which I chalk up to the pandemic making digital notetaking the norm. Between 2019 and 2023 it was popular to say that a set of reasons you were providing were “structural,” even if they were not actually structural causes in the academic sense of the word.
How could you even be disempowered by AI in such a sport dripping with human influence?
Well, humans have been wondering this themselves for quite some time. Harish Natarajan, a world finalist in 2016 and European champion in 2012, participated in an exhibition debate against IBM’s Project Debater in 2019 on the topic “We should subsidize preschool.” Unlike the well-known successes of DeepMind’s narrow AI against human experts, this was a loss for the AI — Natarajan, in calm and controlled demeanor, persuaded more of the audience to change sides than Project Debater did. Support for Project Debater’s side fell from 79% to 62%, while support for Natarajan’s side rose from 13% to 30%. Although a majority of the audience voted the AI over Natarajan as providing more enriching content (58% to 20%).
And to my eyes as a debater and a judge, he ran circles around the AI — he definitely gave less facts, but was able to frame them more convincingly, scaffold them in ways that were richly structured, and more directly argued for his side of the motion than the AI did.
Harish and the machine.
But the story nowadays is quite different. Hackenburg et al. 2026 found that in dyadic text conversations on UK politics, modern (for the time) AIs were much more persuasive than elite debaters. And these debaters were really good, comprising 56 elite debaters, including four world champions and eleven continental champions, being incentivized financially to be persuasive. But they did also find that models performed comparably to coached elite humans when constrained to roughly human level message length and speed, which gives us a more mixed picture (and this mostly depends on what you consider to be “fair comparison”).
Sidenote: you may observe that this plot seems to quite vindicate elite debaters, as being better than laypeople and professional canvassers.
This is a nice conversation through history! In 2019, specialised AI could marshal information but not persuade as effectively as an elite debater. In 2026, general-purpose frontier AI can outperform elite debaters in text persuasion given equal time, but not when given equal speed and message length. So, overall, the evidence for capability is mixed, but trending significantly in favor of AIs as they get more capable, especially as we see more effort poured into not only coding and math capabilities, but also conceptual reasoning.
But what about the community? What do debaters themselves think, feel, and do? We can look at examples from debate events over the years. This is taking some cues from Debating Seriousposting, an international Facebook group for competitive debaters.
Early on there was some interest in using GPT-3 to generate essay outlines for debate motions. The author estimated that only about 30% were reasonable and relevant, but the rest were either obviously incoherent, answered a different topic, or simply restated the motion without adding anything new. This exercise was repeated with ChatGPT, which was much better but still rudimentary, with repetitive and generic content, but clearly relevant, and reactive to both complicated infoslides and material provided during a round. Although, there was already some concern: “What is alarming is the sheer volume of content it is able to produce from minimal effort which also mitigates a lot of strategic considerations in debating.” and some worry about cheating.
Then, in preparation for the World Universities Debating Championships, a debater describes her use of an AI chatbot to prepare a casefile. She and her partner selected questions and topics, to which the model supplied arguments, examples, intuition pumps, and the resulting casefile was “completely written by an AI chatbot.” During a round about a Pacific Alliance common currency, she mentions reading some of the generated material word for word. Here, some cognition is being deferred, but overall the author reports that the AI casefile is “just like any other casefile”: the debater still has to select when and how to use the evidence in the specific debate that is happening. And in the example, the debater still had to read the whole casefile carefully before the competition, checking it for mistakes.
Several things have happened since then:
All the while, AI researchers figure out chain-of-thought, then RLVR, and generally scale up model size and whatnot. Novel progress is made in mathematics, and much of conventional software engineering is successfully automated. By this point, language models have gotten much better, Fable 5, GPT-5.6 Sol, now highly competent reasoners and writers.
Throughout this time, I was also debating, coaching, judging, running tournaments. In debates, I didn’t notice anything out of the ordinary, but in the course of coaching and judging, I would find that some debaters (especially children) would recite obvious AI phrases verbatim, which was a clear red flag. Some debaters would be caught using their phone or laptop in prep time (which was still illegal under tournament rules even before AI), but using it for ChatGPT instead of Google. And as chief adjudicator, many would write their applications for subsidy entirely in AI, which was obvious.
And now, there are governance issues. Major international tournaments have previously collected phones during the elimination rounds to mitigate cheating. But these are later, more prestigious rounds, and the majority of debaters will only be able to do the preliminary rounds, for which it is a logistical nightmare to collect phones. This year, the Australasian regional tournament and the Asian regional tournament in the British parliamentary format implemented a full ban on the use of electronic devices for any purposes during all rounds, and the world championships will be doing the same. To support this, tournament guidance encourages judges to photograph suspected unauthorised tech use, which IMO is quite a powerful image of AI changing the activity even where nobody successfully uses it.
In general, there are recommendations to return to purely paper note-taking and adjudication (to prevent judges from simply transcribing and feeding in a whole round to an AI, and then regurgitating the call), supervised, in-person judge tests, oral interviews about the test rounds, collecting phones throughout all rounds. And maybe giving up on online tournaments entirely, or having separate cyborg and non-cyborg tournaments.
What Now?
You can imagine a world where someone uses AI to: generate every practice case, give feedback after every speech, build AI-generated casefiles, study AI-generated notes from past finals, arrive at Worlds, surrender their phone, and give a completely unaided speech to terrific applause. This doesn’t sound that bad, and this person may have worked quite hard. But I am curious about what goes on in their mind; do they feel empowered, or disempowered?
On one hand, we use technology all the time, such as using the internet to gather data and facts. But we have to resist the urge to be too galaxy-brained here: some crucial part of debating lives in the reasoning, the coming up of arguments. If debate was scored purely on the performance of a speech and the impression it left on the viewer, it would lose something essential that distinguishes it from pure performance arts. Also, some part of it lives in the culture of hanging out with your partner to do a casebuild, to grab a beer after a tournament and not be worried if the guy on the other side is cheating. And some part of it lives in whether or not the debater in question retained valuable critical thinking skills from this process.
In general, how much of the felt value of a sport comes from the (relatively artificial) obstacles? Suits defines a game as the voluntary attempt to overcome unnecessary obstacles. Nobody actually needs to summit Everest without oxygen, or put a basketball into a hoop. But, the pointlessness itself is integral to making the achievement legible. It’s possible that inhaling xenon gas really does increase red blood cell count and improve cardiovascular endurance. In the limit, there is no clear difference between certain drugs and certain tools. A stimulant to help you focus (such as coffee) and a model that provides crucial facts both substitute some amount of capability. Choosing to give 15 minutes of prep rather than 10 is also arbitrary, some formats such as Asian parliamentary do 30 minutes. As such, while there may be differences in degree and character, the line is ultimately down to what the community feels it can tolerate.
Besides, what the debater feels isn’t exactly the right question. The whole point of gradual disempowerment is that the person is making choices that feel locally correct, but are nonetheless objectively disempowered and dependence-cultivating. We could ask it in three different ways:
In light of lacking or mixed conclusions for each of these questions, what should we conclude from the evidence? Well, some things stand out.
Debating is still alive and healthy, as far as I can tell. Tournament attendances haven’t precipitously declined – in the last world championship, we got quite a large crop of judges and debaters to come to Sofia, Bulgaria.
A gaggle of tournament participants at Sofia WUDC 2026. Evidence of vitality.
To rebut doomerism, people will still travel far and wide to sit in badly ventilated rooms and argue about useless things. I don’t think we have to preserve every inconvenience of debate for its own sake, easier access to information is good, cheap feedback for novice debaters in far-flung communities is good. But the writing is on the wall: it’s getting harder and more costly to stop AI cheating, and many aspects that we like about debate are getting outsourced. There is a future looming, of a debate hollowed out and without much substance remaining.
Provisionally, if I had to give a verdict, I’d say:
Perhaps Rooney is right that debate is meaningless. Certainly, most of the motions are forgotten the next morning, between trophies gathering dust and hangovers blearily clearing. But wanting it so much was not meaningless to me: the wanting was what made a shy, bookish boy learn to speak, think, travel. My worry is therefore not really that future debaters will stop wanting to win, but that winning may cease to require them to become the kinds of people who could have done it themselves.
Thank you to Sam Huang whose initial draft I took some evidence and structure from. Thank you to George Wang, Matt Caito, Dan Murfet, Marina Koljic, Nora Flynn-McIver, Stuart Johnson, and Jordyn Gibson for helpful feedback and discussion.