AI Tools for Literature Reviews: Are Current Guidelines Meeting the Needs of Researchers?
Author(s)
Grace E. Fox, PhD1, Nita Santpurkar, MSc2, Juliette Torres Ames, MSc3, Jed Avissar, MSc3, Emanuele Arca`, MSc4;
1OPEN Health HEOR & Market Access, Strategic Market Access, New York, NY, USA, 2OPEN Health HEOR & Market Access, Strategic Market Access, Maharashtra, India, 3OPEN Health HEOR & Market Access, Strategic Market Access, London, United Kingdom, 4OPEN Health HEOR & Market Access, Strategic Market Access, Rotterdam, Netherlands
1OPEN Health HEOR & Market Access, Strategic Market Access, New York, NY, USA, 2OPEN Health HEOR & Market Access, Strategic Market Access, Maharashtra, India, 3OPEN Health HEOR & Market Access, Strategic Market Access, London, United Kingdom, 4OPEN Health HEOR & Market Access, Strategic Market Access, Rotterdam, Netherlands
OBJECTIVES: Artificial intelligence (AI) offers the opportunity to make literature reviews—historically human-driven and resource-intensive processes—more efficient. Unlike systematic literature reviews (SLRs), which are known for their rigor, pragmatic literature reviews have greater methodological flexibility. As such, reviewing and implementing guidelines and recommendations is crucial to ensure that AI is effectively and ethically employed across different types of literature reviews. This study aimed to explore guidelines and recommendations for utilizing AI in literature reviews, including SLRs and pragmatic literature reviews.
METHODS: A scoping review was conducted in January 2025 to identify guidelines and recommendations for using AI in literature reviews. Sources included PubMed; organizations and societies, such as Cochrane and the Centre for Reviews and Dissemination; and working groups, such as Responsible AI in Evidence Synthesis (RAISE). The focus was to identify recommended workflows (e.g., AI-driven, human-in-the-loop) and phases where AI is recommended. Additionally, the extent to which guidelines addressed the use of AI for SLRs compared with pragmatic literature reviews was assessed.
RESULTS: AI is recognized by key stakeholders as a valuable tool for literature reviews, and guidelines and recommendations generally suggest that AI should augment—but not replace—human efforts. The information was predominantly focused on SLRs, where most recommendations were provided for title/abstract screening and data extraction. However, there was less information related to other SLR phases, such as reporting. Gaps were identified in addressing how to adapt AI to the less structured nature of pragmatic literature reviews, where reference sets are generally smaller than for SLRs.
CONCLUSIONS: This scoping review highlights the evolving role of AI in evidence synthesis at a time when many existing guidelines are being updated and new ones are under development. It also highlighted the need for best practices for AI use in pragmatic literature reviews to expand AI’s utility across diverse literature review methodologies.
METHODS: A scoping review was conducted in January 2025 to identify guidelines and recommendations for using AI in literature reviews. Sources included PubMed; organizations and societies, such as Cochrane and the Centre for Reviews and Dissemination; and working groups, such as Responsible AI in Evidence Synthesis (RAISE). The focus was to identify recommended workflows (e.g., AI-driven, human-in-the-loop) and phases where AI is recommended. Additionally, the extent to which guidelines addressed the use of AI for SLRs compared with pragmatic literature reviews was assessed.
RESULTS: AI is recognized by key stakeholders as a valuable tool for literature reviews, and guidelines and recommendations generally suggest that AI should augment—but not replace—human efforts. The information was predominantly focused on SLRs, where most recommendations were provided for title/abstract screening and data extraction. However, there was less information related to other SLR phases, such as reporting. Gaps were identified in addressing how to adapt AI to the less structured nature of pragmatic literature reviews, where reference sets are generally smaller than for SLRs.
CONCLUSIONS: This scoping review highlights the evolving role of AI in evidence synthesis at a time when many existing guidelines are being updated and new ones are under development. It also highlighted the need for best practices for AI use in pragmatic literature reviews to expand AI’s utility across diverse literature review methodologies.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
P24
Topic
Study Approaches
Topic Subcategory
Literature Review & Synthesis
Disease
No Additional Disease & Conditions/Specialized Treatment Areas