Artificial Intelligence (AI) Methods
The use of AI in evidence synthesis is a promising development, provided that tools are applied appropriately and methodological principles of rigor, transparency, and replicability are safeguarded. But not all AI solutions meet these principles. Recent evaluations support this cautious optimism (see BMJ Open: https://doi.org/10.1136/bmjopen-2023-072254).
Across disciplines, there is a growing recognition that AI can support several stages of systematic reviews and other forms of evidence synthesis. At the same time, its constraints need careful consideration. AI may be used only when it aligns with the principles of research integrity. This requires full transparency about when and how it is applied, continuous human oversight, and clear accountability for the final work. Researchers must also explain the rationale for using AI and show that its involvement does not undermine methodological rigor or the trustworthiness of the review.
The NICE position statement outlines expectations for using AI in evidence generation (https://www.nice.org.uk/position-statements/use-of-ai-in-evidence-generation-nice-position-statement).
Cochrane Evidence Synthesis and Methods has highlighted this balance in its special issue on AI in evidence synthesis, bringing together invited articles that both demonstrate the potential of AI and articulate principles for its responsible, methodologically grounded implementation.
https://onlinelibrary.wiley.com/doi/toc/10.1002/(ISSN)2832-9023.AI_in_Evidence_Synthesis
Examples of AI and automation tools used in evidence synthesis
Rayyan
A free web-based tool that accelerates title and abstract screening by allowing reviewers to quickly sort, label, and compare decisions on large sets of citations.
Covidence
An online platform that streamlines study selection, data extraction, and risk-of-bias assessment, automating several workflow steps and supporting collaborative review. Access requires an individual or institutional subscription, though some organisations provide free access.
ASReview
An open-source tool using active learning to prioritise the most relevant records during screening, reducing workload while maintaining high sensitivity.
Abstrackr
An AI-assisted citation-screening platform that predicts relevance and improves with user feedback. A free account is required.
EPPI-Reviewer
EPPI-Reviewer manages references, stores PDFs, and supports both qualitative and quantitative analyses, including meta-analysis and thematic synthesis. It also incorporates newer text-mining features designed to make the systematic review process more efficient.
DistillerSR
DistillerSR uses AI and structured workflows to automate the management of literature collection, screening, and assessment. Whether conducting a traditional systematic review, a rapid review, or a living review, DistillerSR helps streamline the process and produce transparent, audit-ready, and compliant outputs.
Comparison of AI and Automation Tools for Evidence Synthesis
| Tool | Stages Supported | Usability | Performance | Cost | Openness |
|---|---|---|---|---|---|
| Rayyan | Title/abstract screening; deduplication; collaboration | Very easy; intuitive | Fast; team-friendly | Free | Open access |
| Covidence | Screening; full-text; extraction; risk of bias; PRISMA | Easy–moderate | Reliable end-to-end workflows | Payed subscription (free trial, restricted to screening 500 records) | Closed source |
| ASReview | Screening with active learning | Moderate | High sensitivity; reduced workload | Free | Open source |
| Abstrackr | Screening with prediction + prioritization | Easy | Improves with user feedback | Free | Open access (free account) |
| EPPI-Reviewer | Screening; coding; thematic synthesis; meta-analysis; text mining | Moderate | Very powerful; complex projects | Paid subscription (one month free trial) | Open source |
| DistillerSR | Screening; extraction; audit trails; AI suggestions | Easy | Strong for audits & large teams | Paid subscription | Closed source |
Moving forward, the challenge is not whether we use AI in evidence synthesis, but how we do so responsibly. Our team will continue to explore, test, and share best practices as this landscape evolves. We also welcome insights from colleagues—whether experiences, questions, or examples of integrating AI into review workflows. Feel free to share your reflections with us at esca@amsterdamumc.nl.