OBJECTIVES: Systematic literature reviews (SLRs) are critical for synthesizing evidence in healthcare, social sciences, and other research fields. However, generating and optimizing search strategies for these reviews can be time-consuming and labor-intensive, requiring expert knowledge to ensure comprehensive and unbiased results. This research aimed to develop an innovative tool to generate search queries for SLRs using generative AI.

METHODS: A user interface was developed using Python 3.10, GPT-4o, and the Flask framework to analyze existing literature, identify relevant keywords, and suggest optimal combinations of search terms. The application allows users to start their search by selecting a literature medical database (PubMed, etc.), followed by the specification of PICOS criteria, detailing population, intervention, comparator, outcomes, and study design for their research problem. The effectiveness of the tool was assessed through a comparative analysis with the traditional manual method, focusing on metrics such as recall, precision, and overall time efficiency.

RESULTS: The tool was tested to generate the search query for the PubMed database for an SLR focused on chronic rhinosinusitis with nasal polyps. A domain expert extensively examined and adjusted search queries generated by the tool. The AI-powered searches were comparable to manually generated ones in terms of both comprehensiveness and efficiency. The AI tool significantly reduced the time required to develop and refine the search strategy, allowing researchers to focus more on data analysis and interpretation.

CONCLUSIONS: The AI tool using GPT-4o considerably enhances search query generation for the SLRs in medical research, speeding up the process and improving the precision and comprehensiveness of literature searches. This research advocates for the broader adoption of AI technologies to transform evidence synthesis and support robust, data-driven decision-making