A Systematic Review on Explainable Transformers and the Evolution of E-commerce Recommendation Systems

Artificial intelligence (AI) has played a critical role in the development and operation of e commercerecommendation systems, causing an evolution in systems design and functioning to more complex systems that leverage deep learning and generative AI. This systematic review critically reviews a broad range of recommendation models—starting with the clas sical methods of collaborative filtering and content-based filtering, neural networks, and transformer-based architecture models. Special emphasis is given to explainable transform ers, which exploit attention mechanisms to track subtle user-item interactions to be able to provide contextual personalization in dynamic online settings. The review follows the PRISMA guidelines and considers peer-reviewed publications by scholars within the period dating back to 2016 to 2025, and retrieved in such key scholarly sources as IEEE Xplore, ACM Digital Library, SpringerLink, ScienceDirect, Scopus, and Web of Science. Important benchmarks against which performance will be assessed involve predictive accuracy, scalability, and cold-startability, and model explainability. Through organizational and synthetical analysis of the current status of the field, the review provides a compacted source of information both to scholars and practitioners who need to learn about the possibilities and restraints of modern recommendation systems