ReviewAid: An Open-Source Tool for Efficient PICO-Based Screening and Data Extraction in Systematic Reviews

Systematic reviews are the gold standard for evidence synthesis, yet manual screening and dataextraction remain significant bottlenecks. This paper presents ReviewAid, an open-source, AI-powered software tool designed to streamline this workflow. Built on Python and Streamlit andpowered by the GLM-4.6V-Flash multimodal model, ReviewAid facilitates PICO-basedscreening and customizable data extraction. To address reliability challenges, the softwareintroduces a Hierarchical Four-Tier Confidence Scoring System and a Bulletproof ParsingPipeline capable of recovering data from malformed API responses. Featuring a real-time SystemTerminal for operational transparency, validation on 100+ articles demonstrated high processingspeeds and robust error handling. ReviewAid is architected not as a replacement for humanjudgment, but as a "third reference" layer that accelerates the review process under an Apache 2.0License.