Microsoft Copilot Researcher uses two AIs to critique research reports

Microsoft Copilot Researcher introduces Critique, running two AI models in parallel—one for planning, sourcing, and drafting reports, the other for reviewing accuracy and completeness. It becomes default in Auto mode and outperforms single-model systems by 7 points on 100 complex tasks, plus beats Perplexity by 14%. Council tool compares multiple models side-by-side with agreement summaries. This addresses gaps in evidence, source quality, and topic coverage from single-model research.