Argue with an AI.
Become a sharper thinker.

No. The default mode is called Quick Clash — Pro vs. Con on any topic, no rules, no jargon, no speech order. You type a topic, pick a side, the AI takes the other side, and you trade arguments turn by turn.

If you've never debated before, that's the mode the app is built around. Competitive formats (APDA, BP, Public Forum, etc.) are available too, but they're opt-in — you won't be dropped into one by accident.

Real things real users do:

Argue better at work — stress-test a pitch before a meeting. Catch the hole in your own thinking before a manager does. Prep for a tough conversation with a friend or partner. Practice defending a class presentation. Win (or honestly lose) an argument in a group chat. Think through a policy question you care about and find out where your position actually breaks.

If you do compete, the app also handles tournament prep — case gen, round sim, judge analysis, argument mapping, the full stack.

Click Debate Chat in the top nav. You'll see a "Random motion + background" button — one click fills in a topic and the context. Pick Pro or Con. Hit Start. Type your first argument. That's the whole setup.

If you want to learn the ropes before jumping in, the Guide tab has a 60-second intro for non-debaters, plus the deep version for people going to tournaments.

ChatGPT agrees with you. This doesn't — the AI is explicitly instructed to take the opposite side and pressure-test your reasoning. It finds the weak link, hits it, and forces you to defend or update your view.

The prompts also encode how actual competitive debaters argue: named historical anchors (not "studies show"), explicit concessions before attacks, multi-link warrant chains, weighing on time-frame and logical priority, and format-specific moves (two-voter PMR collapse in APDA, extension discipline in BP, etc.).

You're not getting a chatbot that happens to know about debate. You're getting a debater.

Four models, and you can pick: Claude (surgical and structured), GPT (creative and rhetorical), Gemini (fast and well-rounded), Grok (unfiltered and sharp). Each argues differently; the difference is noticeable after a few rounds.

Some features default to a specific model because it's strongest at that task (e.g. case generation defaults to Claude). You can override in any feature with a model toggle.

Yes, if you sign in. Click Teach AI your style in the top bar and fill out an 8-question questionnaire: experience level, preferred format, argument style, jargon tolerance, debaters you admire, things you want to avoid. Plus the app quietly learns from how you actually write in Debate Chat (jargon level, sentence length, how often you concede, whether you cite named examples).

That profile gets mixed into every AI response so the AI matches your voice instead of forcing you to match its. If you stay anonymous, every session starts fresh — the style profile lives with your account.

Push back. Seriously — the AI is a debate opponent, not an oracle. If it cites a fact that's wrong or makes a logical leap, call it out. That's exactly the kind of move a strong debater should be able to handle, and it makes you sharper.

If the AI's output is just bad (repetitive, off-topic, evasive), hit the feedback link in the beta banner at the top of the page. That's how we tune the prompts. Every round of feedback has changed something in the system.

Free during beta. Every tool, every AI model, no limits, no credit card, no sign-up required.

Anonymous users get 5 requests before a feedback form opens — fill it out and you unlock unlimited use immediately. Signed-in users skip that step because signing in counts as engagement we can follow up on.

After beta, pricing will be clearly announced before anything changes. Users who gave feedback during beta keep unlimited access.

Yes. It's a web app — works in any browser, nothing to install. On iOS and Android you can also add the site to your home screen for an app-like experience (offline shell, proper icon, full-screen).

Voice debate mode requires microphone access and works best with good headphones, especially for long rounds.

Your Debate Chat rounds, saved cases, and style profile are stored in your browser's localStorage by default — nothing leaves your device unless you sign in.

If you sign in, your saved cases and style profile sync to your account so they follow you across devices. Message content is sent to the relevant AI provider (Claude/GPT/Gemini/Grok) to generate responses; we don't train models on your content and we don't sell data.

You can delete any saved round from the My rounds library in Debate Chat at any time.

Yes — the BYOK (bring your own key) flow lets you paste an Anthropic API key into the app and run unlimited requests billed directly to your own account. Useful if you hit rate limits, want a specific model version, or just prefer not to rely on our infrastructure.

Keys stay in your browser (localStorage) and are never sent to our servers. We don't see them.

APDA Parliamentary, British Parliamentary, World Schools, Public Forum, Lincoln–Douglas, Policy, Congressional Debate, plus Quick Clash for informal practice.

Each format gets proper speech order, timers, POIs, reply speeches where applicable, and a judge ballot at the end. Case Gen output is structured for the format you pick (PMR collapse for APDA, extension for BP, contention-based for PF, etc.). Prompts are format-scoped — the AI argues differently in NPDA vs. APDA.

Debate AI is a full timed voice round — you deliver your speeches out loud, the AI delivers its speeches as a voice opponent, POIs happen mid-speech, and a judge ballot comes out at the end. It's the closest simulation of a real tournament round.

Debate Chat is a text-based back-and-forth. No timer, no speech structure, just trade arguments until you type "end" for a verdict. Lower stakes, faster iteration, easier to study a specific clash without running the whole round.

Most people use Chat for prep and Debate AI for simulation.

No. It's a prep tool — same category as reading an article, running a practice round with a teammate, asking a coach, or looking through old case lists. You're still the one speaking and thinking in the actual round.

No circuit we're aware of prohibits using AI tools in prep. Using AI during a live round is obviously different and would be cheating in any format. The app explicitly separates prep features (case gen, mapper, chat) from in-round features (none — by design).

At the end of a Debate AI round (or when you type "end" in Debate Chat), the app generates a full RFD: who won, why, the central clash that decided the round, speaker points for each speech, specific strengths, specific things to improve, and one drill to practice before your next round.

The ballot is prompted to judge like a circuit judge, not like a kind teacher — expect honest feedback. You can override the brain used for judging (e.g. judge with Claude even if you debated with GPT) for a second opinion.

Yes. Sign in, create a team, invite teammates by email. Your saved cases become shareable with the team from the Saved Cases library — team members can import any shared case into their own saved list.

Team plans also unlock shared analytics so you can see what your squad is working on, and team-wide usage limits that don't eat into individual quotas.