Abstract
<title>Abstract</title> <p> <italic>Introduction</italic> : Pandemic vaccination requires rapid and systematic post-marketing safety monitoring, particularly to detect rare, complex adverse events of special interest (AESI) that are often not identified during early clinical trials. Artificial intelligence has the potential to strengthen active vaccine safety surveillance by enabling the large-scale use of both structured and unstructured data. Objective: This scoping review aims to map the evidence on the use of AI in active surveillance for AESI following pandemic vaccination, particularly COVID-19 and influenza vaccines. <italic>Methods</italic> : This scoping review was conducted in accordance with the PRISMA-ScR guidelines. Literature searches were performed in PubMed, Ovid-Embase, Web of Science, Scopus, Google Scholar, and through hand searching up to 30 April 2026. Eligible studies were primary studies that applied AI methods, including machine learning, deep learning, natural language processing, or hybrid models, in active vaccine safety surveillance to detect AESI or health outcomes of interest following pandemic vaccination. Data were narratively extracted according to study characteristics, data sources, AI methods, AI roles, safety outcomes, implications, and limitations. <italic>Results</italic> : Of 20,098 records identified, 20 studies met the inclusion criteria. Most studies were published in 2021–2022, with 13 studies (65.0%), and the majority focused on COVID-19 vaccines, with 17 studies (85.0%). The most frequently investigated outcomes were multiple or broad AESI and maternal outcomes. The most common data source was electronic health records, including clinical notes and imaging reports, used in 9 studies (45.0%), followed by social media or online platforms in 6 studies (30.0%), as well as claims databases, online surveys, and physiological monitoring data. AI was used for case detection, extraction of adverse events from free text, risk prediction, signal screening, social listening, and early signal exploration. Its role remained predominantly supportive for signal detection and prioritisation. <italic>Conclusion</italic> : This review demonstrates that AI enhances active vaccine safety surveillance through dual case detection and strategic support. By leveraging diverse health data, AI accelerates candidate identification, clinical information extraction, and early signal prioritisation of AESI. These findings strengthen global post-marketing surveillance frameworks, guiding future research towards optimal and reliable AI implementation. </p>