Visual search technologies have become pivotal in enhancing online shopping experiences, particularly in second-hand markets, where accurate text descriptions are often limited or inconsistent. The proposed system, Snap & Shop, is a web-based platform designed to enable users to find similar products across three popular online marketplaces serving the Portuguese market. The motivation arose from the absence of integrated solutions combining image-based search with aggregated, relevant results tailored to the Portuguese second-hand sector. To achieve this, the system employs machine learning models alongside web scraping techniques for dynamic data collection. At its core, the system utilizes a multimodal Large Language Model (LLM) with a custom prompt for product identification, coupled with a classification model that extracts visual features from images, which are then compared to assess their similarity. Data acquisition is performed via web scraping, using simulated browser behavior through undocumented yet publicly available platform-specific endpoints to bypass restrictions without requiring a dedicated database. This method allows real-time retrieval of relevant product listings from multiple marketplaces. Experimental results demonstrate the system’s effectiveness in retrieving highly relevant, visually similar products, highlighting the potential of visual search within the Portuguese used goods market. The integration of advanced machine learning models with web scraping techniques proves to be a scalable and efficient approach to enhance product discovery. This research establishes a foundation for future developments, including improved search precision, platform interoperability, and sophisticated marketplace functionalities, ultimately contributing to a richer, more accessible online second-hand shopping experience.