DebateAI coaches your reasoning through live AI debate and structured feedback. Turn argumentation into a trainable skill.
Built for structured argument practice, debate training, and reasoning improvement.
Product Preview
Not a chatbot. Not a quiz. A structured turn-based debate with an AI opponent that challenges every argument you make.
Two Modes
A turn-based debate against an AI opponent. Every argument is scored. Every session ends in a verdict. This is structured intellectual competition.
Open-ended practice with adaptive AI coaching. No fixed turn pressure. The AI guides you toward stronger arguments, then tells you exactly what to work on next.
Philosophy
Argumentation is a skill — but most people treat it as a personality trait. When someone is wrong in a meeting, a debate, or a negotiation, the response is usually louder, not sharper. DebateAI was built on a different premise: that reasoning can be trained, measured, and improved, just like any other cognitive skill. The Arena is the practice ground.
“The quality of your arguments determines the quality of your decisions. DebateAI turns argumentation into something you can actually work on.”
Process
A complete session takes 10–20 minutes. The report lasts much longer.
Enter any debate topic or choose from curated categories: ethics, policy, technology, philosophy, economics. Set your stance: For or Against.
Challenger for competitive scored debate. Coach for guided open practice. Set difficulty, archetype (Challenger), and coaching style (Coach).
Argue turn by turn against your AI opponent or coach. Request a hint. Watch your score shift in real time. Every argument is evaluated.
A structured report appears the moment the session ends: verdict, score, missed opportunities, logical fallacies identified, and your improvement path.
Challenger Mode
Challenger is turn-based, time-pressured, and competitive. You pick your parameters before entering, and the AI adapts to play against your chosen stance with precision and resistance.
Six archetypes available: Socratic (questions your premises), Devil's Advocate (steelmans the opposition), Empiricist (demands evidence), Rhetorician (challenges your framing), Iron Fist (relentless pressure), Mirror (forces self-examination).
Coach Mode
Coach Mode is your private practice room. No fixed turns, no score pressure — just you, a patient AI, and a continuous feedback loop that identifies exactly where your reasoning breaks down.
Coach Mode adapts to the patterns in your arguments. If you frequently overgeneralize, it will surface that. If you avoid evidence, it will push you there. Every session ends with a personalized improvement plan — not a score.
Post-Session Reports
Every session generates a structured report. Challenger reports deliver verdicts and fallacy analysis. Coach reports deliver coaching moments and growth plans.
“You held a consistent position and demonstrated structural clarity, but your opponent successfully exploited two underdeveloped premises in turns 2 and 4.”
“Your analogy use is improving. The core weakness this session was not extending your framings before moving to the next argument. Depth before breadth.”
Dashboard
Every session is saved. Progress compounds. The dashboard shows where you’ve been and where to go next.
Architecture
DebateAI is a production-grade AI application, not a demo. Every component is structurally sound, API-backed, and state-consistent.
Every debate session is driven by structured prompts with mode-specific system contexts. Challenger and Coach have separate prompt chains and evaluation schemas.
Session transcripts and reports are stored server-side. History is retrievable. The Report Detail view is the same data as the live session — no loss of fidelity.
Challenger and Coach reports use different schemas. Challenger: score, verdict, metric breakdown, fallacy list, turn timeline. Coach: quality ring, coaching moments, next drill.
System prompts enforce role separation. The AI opponent cannot break character. The coach cannot score like a challenger. Mode boundaries are enforced at the prompt layer.
React + TypeScript frontend. Route structure covers 17 pages. Component families for Arena, Report, Setup, Dashboard, and Auth are independently composable.
Token-based color system. Scholar's Arena palette with strict Challenger/Coach mode separation. Lora for editorial moments only. No purple, no cold SaaS blue.
A 3-turn Challenger session, live against the AI. You’ll see the transcript, the scoring, and the post-session report — the full experience in under 5 minutes.
Guest sessions are not saved. Create an account to track your progress.
Every conversation, meeting, and debate is a test of how well you think. DebateAI gives you a structured place to practice.
Free to start · No credit card · Sessions saved permanently