Deep Research: how Arcana handles complex multi-step questions

When a single search is not enough

Most questions have a single clear answer. But some questions require synthesizing information across multiple independent dimensions before a coherent answer is possible. Competitive landscape analysis. Academic literature reviews. Policy research that spans multiple jurisdictions. Questions that require understanding not just what is true, but why, and what it implies.

For these questions, a single search-and-synthesize pass is insufficient. Deep Research mode was built to handle them.

How Deep Research works

When you submit a complex question to Deep Research, Arcana first decomposes it into a set of sub-questions that collectively cover the problem space. Each sub-question is then researched independently, with its own source retrieval, synthesis, and confidence scoring. Finally, the individual findings are merged into a unified report that addresses the original question in full.

The entire process typically takes 30 to 90 seconds, depending on the complexity and breadth of the question.

What the output looks like

Deep Research returns a structured report with clearly labeled sections, inline citations for every claim, and a source list at the end. Where sub-questions yielded conflicting findings, the report surfaces the disagreement explicitly rather than silently resolving it in favor of the majority view.

Appropriate use cases

Deep Research is most valuable when the question genuinely requires breadth: market research, due diligence, academic background research, competitive intelligence, or any situation where you need a comprehensive picture rather than a single fact. For simple factual queries, standard search is faster and equally accurate.

Availability

Deep Research is available on Pro and Enterprise plans.