Artur Nowak represented Evidence Prime at the GRADE AI Project Group Workshop in Porto, Portugal, on 18–19 November 2025. Hosted at the Faculty of Medicine of the University of Porto, the meeting examined how AI could support guideline development and evidence synthesis within the GRADE approach. The Porto Cochrane Centre’s account describes the workshop’s focus on the relevance, strengths and limitations of these technologies.
Summarising evidence for guideline decisions
On 18 November, Artur and Ignacio Neumann co-led the session “Evidence-to-decision framework: Summarising evidence for each criterion.” This focused our contribution on a practical part of guideline development: preparing evidence summaries that help a panel make and document its judgments.
Retrieval and summarisation need to be considered together. A readable summary is useful only if it draws on the evidence relevant to the question. Reviewers also need to be able to check where statements came from, understand what was left out, and identify uncertainty that could affect the discussion.
For us, this is an important distinction when considering AI assistance. Drafting text can reduce one part of the work, while deciding whether that text is an adequate and balanced account of the evidence remains a methodological task. Supporting the panel requires attention to both.
Connecting individual tasks with the guideline process
The wider programme considered AI across several stages of guideline development, including question and outcome prioritisation, assessment of certainty, adaptation to different contexts, and communication of recommendations. The organiser’s workshop overview also identifies prototype tools and support for people applying GRADE among the topics discussed.
The meeting provided a setting to examine these tasks together. Our contribution brought an evidence-synthesis technology perspective to the discussion of how summaries can support transparent guideline decisions, with methodological review and panel judgment remaining central.
