Consensable works by running your question through several leading commercial AI models in parallel, such as Claude, GPT, Gemini, and Sonar. Each of these panel members performs its own real-time web search before answering. A chair model then synthesizes these individual responses into a single, accountable answer, complete with a panel-agreement score that indicates how much the models converged on the key claims.

The engine behind Consensable is a multi-AI synthesis system. Unlike a single model, this approach allows the product to surface explicit disagreements between models and provide a panel-agreement score. The synthesis process typically takes between 15 and 60 seconds, depending on whether you choose the fast, balanced, or thinking preset.

Regarding costs, Consensable operates on a credit-based system. You use credits to submit queries, as each query incurs costs based on the provider AI charges for every model call involved in the synthesis. This includes the initial responses from each panel member, the chair's synthesis pass, and any dispute rounds.

Queries that cost more credits generally involve longer or more complex questions, the use of larger or premium models, more dispute rounds (such as those triggered by the balanced or thinking presets), and larger context windows resulting from live web searches and cited URLs. Reading a result that has already been generated and shared is always free, and using the /help mode is also free.