AI in Systematic Reviews: What the RAISE Recommendations Mean for Researchers

AI tools promise to accelerate systematic reviews, from screening publications to extracting data and drafting summaries. However, producing a review quickly does not establish that its methods are reliable or its conclusions justified.

The Responsible use of AI in evidence SynthEsis (RAISE) initiative provides recommendations for responsible practice, alongside guidance on developing, evaluating and selecting AI tools. Its central principles include transparency, human oversight and continued author accountability for the methods and findings of a review.

For review authors, this means evaluating whether a tool is appropriate for the specific task and evidence base. Performance demonstrated in one clinical area or publication type may not transfer to another. The justification for using AI should therefore consider its validation, limitations and potential consequences of error. Where AI makes or suggests judgements, its contribution should be reported transparently.

In practice, researchers should ask:

• Screening: Could the tool exclude eligible studies, and how would omissions be detected?
• Data extraction: Can each estimate be traced to the correct population, outcome, time point and analysis?
• Risk of bias: Are judgements supported by the study report, or merely expressed convincingly?
• Synthesis: Are the studies sufficiently comparable to justify pooling, and do the conclusions reflect uncertainty?

These questions require methodological understanding. A plausible explanation of bias is insufficient if it misrepresents the study; an accurately extracted estimate may still be inappropriate for the planned meta-analysis.

RAISE provides a useful framework for examining AI-assisted workflows critically. For researchers, the practical implication is clear: efficiency gains must be evaluated alongside the trustworthiness of the resulting evidence. Critical appraisal skills remain essential for recognising errors, challenging assumptions and defending the final conclusions.

Read the guidance: RAISE recommendations and supporting resources.

To discuss bespoke training in systematic reviews, meta-analysis and critical appraisal of AI-assisted research, contact office@criticalappraisal.com.