Systematic reviews typically follow the PRISMA framework, which involves screening many scholarly works and then extracting evidence from those included. ReviewPilot, described in a new arXiv preprint, is a multi-agent system built around large language models to assist with these two stages.

The system is positioned alongside "deep research" tools that can already search and summarize literature. The authors propose that a multi-agent setup can handle the structured tasks of screening and data extraction, rather than relying on a single LLM pass.

The abstract does not provide experimental results or comparisons, so the claims are limited to the system's design and intended workflow. As a preprint, it has not yet undergone peer review.