Generative artificial intelligence (GenAI) is the new electricity – it’s come with the promise of improving everything we do, and if you can read the writing on the wall, it’s here to stay. Our little corner of science in evidence synthesis and systematic literature reviews (SLRs) has not been untouched by its tentacles. Traditionally expert-driven, structured, thorough research methods have been called into question owing to their time-consuming, labour-intensive, and highly repetitive nature. Is there truly no room for modern development, many asked? At the very heart of evidence synthesis is a need for rigour, transparency, and replicability. And why should this not be the case, when synthesised results are used to inform decision-making by policymakers, healthcare institutions, clinicians, and research institutions, and ultimately to serve the public interest.
A variety of GenAI tools can be used during the evidence synthesis process to assist with search strategy development, record identification, deduplication, data screening, data extraction, data synthesis, or data reporting. The pace of development has been lightning fast, leaving health technology assessment agencies, evidence synthesis organisations, and guideline bodies scrambling to keep astride. With the pace of GenAI tool development outpacing validation studies, researchers and methodologists were alerted of the need for the development of a one-for-all best practice guideline.
The Responsible AI in Evidence Synthesis (RAISE) guidelines are a collaborative initiative developed by Cochrane and partner organisations to provide tailored recommendations for the ecosystem of stakeholders involved in evidence synthesis. On 2nd June 2025, the finalised RAISE guidance and recommendations were published, structured into three main documents. In a webinar, Cochrane’s Evidence Reviews Principal Investigator, Prof. James Thomas, discussed the new best practices.
- RAISE 1 provides recommendations for the eight stakeholders that play a role in the AI- Evidence Synthesis ecosystem.
Recommendations for the evidence synthesist/systematic reviewer:
- The reviewer remains ultimately responsible for the evidence synthesis
- The use of GenAI should be transparently reported
- Name, version, and dates
- Purpose of GenAI tool and impact on the synthesis
- Justification: justify use of tool and method of use, cite tool evaluation studies considering strengths and limitations
- Declare any financial and non-financial interests in the GenAI system or tool
- Ensure ethical, legal, and regulatory standards are adhered to when using AI. Consider the ethical issues of the tool developers when selecting the tool
- Contribute learnings from your experience of the tool to the ecosystem
- RAISE 2 provides guidance on building and evaluating GenAI tools, i.e., aspects of validation, performance metrics, and transparent reporting.
- RAISE 3 assesses the suitability of GenAI tools for specific tasks, i.e., considering ethical, legal, and regulatory factors, and factors to consider when selecting and using GenAI tools.
As the RAISE guidelines start to be implemented and we begin to navigate the new era of GenAI in evidence synthesis, careful consideration is required to balance the obvious utility of the new technology in improving efficiency with the need to demonstrate trustworthiness and reliability and always maintain best ethical practices. The use of GenAI must always be fundamentally methodologically sound, and well justified. The publication of new guidelines provides a basis on which to start putting GenAI into practice, but the use of human discretion will always remain key. We have a responsibility to use GenAI, but enhance, not replace human judgement.
May the Force be with you!
