AI-driven insights into electronic Word-of-Mouth and repurchase intention: A Neural Network approach in multisectoral contexts

Abstract

Electronic word-of-mouth (E-WOM) is widely associated with consumer trust and repurchase intention, but evidence on how artificial intelligence (AI) and neural networks analyse these relationships remains fragmented across industries. This PRISMA-guided systematic literature review synthesizes research across eight sectors: healthcare, banking, tourism, education, insurance, real estate, entertainment, and food and beverages. The review examines sentiment analysis, repurchase intention, customer loyalty, and the fit between neural-network architecture and data structure. Searches identified 3,280 records; 230 studies were included in the qualitative synthesis, and 68 studies reporting quantitative model information were examined descriptively. This is a review of published studies rather than primary model training, and no pooled meta-analysis was conducted because datasets, outcomes, and validation procedures were heterogeneous. The synthesis indicates that contextual language models fit review text, graph neural networks fit relational or coordinated-review signals, multimodal networks require visual evidence, and reinforcement learning fits sequential recommendation settings. Cross-sector conclusions therefore emphasize recurring mechanisms and architecture-to-data fit rather than unsupported comparisons of accuracy values from non-equivalent studies.