How do debates work under the hood?

When you start a debate, AskVerdict:

  1. Assigns specialized roles to AI agents (Supporter, Critic, Risk Analyst, Cross-Examiner)
  2. Each agent independently builds its case based on your question and any context you provided
  3. Agents exchange arguments and counter-arguments across 1-3 rounds depending on mode
  4. A final synthesis agent reads all arguments and produces the structured verdict

All of this happens in parallel where possible, which is why debates stream progressively rather than all at once.

When should I use Fast vs Balanced vs Thorough?

Fast is ideal when you need a quick second opinion: should I send this email, is this vendor quote reasonable, is this a good time to book travel.

Balanced suits most real decisions: career moves, product decisions, evaluating a proposal, choosing between two tools.

Thorough is worth the extra time and credits for high-stakes choices: signing a contract, major financial decisions, architectural choices for large systems, or any decision where you want maximum adversarial scrutiny.

How do I interpret the verdict and confidence score?

The confidence score reflects how much the debating agents agreed after cross-examination:

  • 80-100: Strong consensus. The answer is fairly clear given the information provided
  • 60-79: Moderate consensus. Lean in one direction but meaningful caveats exist
  • 40-59: Divided. The agents found roughly equal merit on both sides; the decision hinges on your specific priorities
  • Below 40: Genuinely unclear. More context or expert input is recommended

A low score is not a failure. It means the question is genuinely complex, which is itself useful information.

How many credits does each mode cost?

  • Fast mode: 1 credit per debate
  • Balanced mode: 3 credits per debate
  • Thorough mode: 8 credits per debate

BYOK subscribers ($19/month) use their own API keys and are not charged credits regardless of mode. Credits are only consumed when AskVerdict's hosted API keys are used.

Can I give the agents extra context?

Yes. The debate input includes an optional context field where you can paste relevant information: a job description, a contract clause, financial data, or background on your situation.

Agents incorporate this context into their arguments. The more specific and accurate your context, the more grounded the verdict will be.

Can I share a debate with someone else?

Yes. Completed debates can be shared via a public link. Go to any completed debate and click Share. You can make the debate public (anyone with the link can view) or keep it private.

Public debates are read-only for viewers. They can see the full argument tree and verdict but cannot run new debates from the link.

What are debate templates?

Templates are pre-structured debate formats for common decision types, such as comparing two job offers, evaluating a startup idea, choosing a tech stack, reviewing a contract, and more.

Templates pre-fill the question format and suggest relevant context fields, so you get faster, more targeted verdicts. Browse the Template Library from the sidebar.

Can I re-run or tweak a past debate?

Yes. From any past debate, click Re-run to start a new debate pre-filled with the same question and context. You can edit before running.

This is useful when you've gathered new information or want to try a different mode on the same question. Each re-run creates a new debate entry and consumes credits accordingly.

Related help topics

Browse adjacent support areas for faster troubleshooting and onboarding.

For production-impacting issues, check the status page first, then include relevant request IDs when contacting support.