Artificial Intelligence (AI) Readiness to Support Evidence Synthesis by Workflow: Findings From a Review of Reviews

Area of Science:

• Information Science
• Biomedical Informatics
• Artificial Intelligence

Background:

• Evidence synthesis is vital for evidence-based practice but is resource-intensive and slow.
• Integrating automation and artificial intelligence (AI) into evidence synthesis can improve efficiency.
• Standardized adoption of AI in evidence synthesis is still developing.

Purpose of the Study:

• To identify peer-reviewed evidence on AI readiness for evidence synthesis.
• To evaluate the capabilities of AI-powered tools in evidence synthesis workflows.

Main Methods:

• A review of reviews was conducted, searching MEDLINE, Embase, and Global Index Medicus.
• Included studies published between January 2020 and May 2025, focusing on evidence synthesis tools.
• Extracted tool features and performance metrics across evidence synthesis stages: search, screening, appraisal, extraction, and synthesis.

Main Results:

• 21 studies reviewed identified 46 evidence synthesis tools.
• Nine tools supported all five evidence synthesis workflow stages; DistillerSR had the most features (19/21).
• AI-powered title/abstract screening tools demonstrated $\ge 95\%$ sensitivity, with EPPI-Reviewer, Research Screener, and SWIFT-Active Screener consistently meeting this threshold.

Conclusions:

• Peer-reviewed evidence supports AI readiness for human-supervised automation of title/abstract screening.
• Evidence for AI readiness in other evidence synthesis tasks is limited.
• AI, particularly in screening (e.g., DistillerSR, EPPI-Reviewer), can enhance efficiency and sensitivity, potentially enabling living evidence syntheses.