An Update on the Evolving Use of Artificial intelligence and Machine Learning (AI/ML) in Systematic Literature Reviews (SLRs)
Author(s)
Sheily Kamra, MBA1, Loveleen Ak, M.Pharmacy2, Deepali Moon, M.Pharmacy3, Anshu Rajput, M.Pharmacy3.
1Sr. Consultant, IQVIA, Gurugram, India, 2IQVIA, Gurgaon, India, 3IQVIA, Gurugram, India.
1Sr. Consultant, IQVIA, Gurugram, India, 2IQVIA, Gurgaon, India, 3IQVIA, Gurugram, India.
OBJECTIVES: Extending on our previous work on the use of AI in SLRs, this review aimed to evaluate the functionalities of AI/ML-enabled web-based and software tools across the SLR workflow, with a particular focus on their application in data extraction, quality assessment, and reporting.
METHODS: Targeted literature searches were conducted on June 7, 2025, using Embase®, MEDLINE®, and Cochrane databases via OVID SP® to identify AI/ML-based SLRs published from 2023 onwards. Search terms included “AI,” “ML,” “deep learning,” “SLR,” “meta-analysis,” and specific AI/ML-enabled platforms. No restrictions were applied on indication, treatment or geography.
RESULTS: Twenty-five SLRs using AI/ML-assisted tools were identified, with a substantial increase in AI/ML tool usage observed over the past two years (3 in 2023, 11 in 2025). The most frequently utilised tools were Covidence (n=9), ChatGPT (n=6), Rayyan (n=5), DistillerSR (n=2) and Nested Knowledge (n=2). Other AI/ML tools identified were Laser AI, Research Screener and Llama2-13b, and Llama3-8b. Among these identified SLRs, AI/ML was predominantly used in data extraction (n=20), followed by screening (n=10) and quality assessment (n=7), with Covidence, ChatGPT, and Rayyan being the most utilised tools for these steps. A limited number of SLRs reported the use of AI/ML for report drafting (n=3) and search strategy development (n=2), all of which utilised ChatGPT. Despite their growing utility, these tools exhibited several limitations, including restricted interpretability, reliance on human oversight, and issues related to integration and computational resource requirements.
CONCLUSIONS: Use of AI/ML in SLRs has evolved rapidly, particularly in data extraction and screening. However, its application remains limited in tasks such as quality assessment, report drafting, and search strategy development. These findings suggested that AI/ML tools cannot reliably be used independently to conduct end-to-end SLRs, and there is a critical need to combine automation with human expertise to ensure methodological rigor and transparency in evidence synthesis.
METHODS: Targeted literature searches were conducted on June 7, 2025, using Embase®, MEDLINE®, and Cochrane databases via OVID SP® to identify AI/ML-based SLRs published from 2023 onwards. Search terms included “AI,” “ML,” “deep learning,” “SLR,” “meta-analysis,” and specific AI/ML-enabled platforms. No restrictions were applied on indication, treatment or geography.
RESULTS: Twenty-five SLRs using AI/ML-assisted tools were identified, with a substantial increase in AI/ML tool usage observed over the past two years (3 in 2023, 11 in 2025). The most frequently utilised tools were Covidence (n=9), ChatGPT (n=6), Rayyan (n=5), DistillerSR (n=2) and Nested Knowledge (n=2). Other AI/ML tools identified were Laser AI, Research Screener and Llama2-13b, and Llama3-8b. Among these identified SLRs, AI/ML was predominantly used in data extraction (n=20), followed by screening (n=10) and quality assessment (n=7), with Covidence, ChatGPT, and Rayyan being the most utilised tools for these steps. A limited number of SLRs reported the use of AI/ML for report drafting (n=3) and search strategy development (n=2), all of which utilised ChatGPT. Despite their growing utility, these tools exhibited several limitations, including restricted interpretability, reliance on human oversight, and issues related to integration and computational resource requirements.
CONCLUSIONS: Use of AI/ML in SLRs has evolved rapidly, particularly in data extraction and screening. However, its application remains limited in tasks such as quality assessment, report drafting, and search strategy development. These findings suggested that AI/ML tools cannot reliably be used independently to conduct end-to-end SLRs, and there is a critical need to combine automation with human expertise to ensure methodological rigor and transparency in evidence synthesis.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
MSR26
Topic
Methodological & Statistical Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics
Disease
No Additional Disease & Conditions/Specialized Treatment Areas